Stage 3: улучшена логика жизненного цикла и очищены ответы ассистента
This commit is contained in:
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval--DLjm5dCSP",
|
||||
"timestamp": "2026-03-26T14:59:09.726Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "dI-qZZYf2cyipx",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "2xKFo8M6aNJ1pp",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "IAo0hvXWAhLedk",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval--lulgUEKkp",
|
||||
"timestamp": "2026-03-26T15:04:48.691Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "7JgPan7VWzpUAo",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "kGlyXyRKB0Ktr5",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-3t1L3QY0wE",
|
||||
"timestamp": "2026-03-26T14:29:33.451Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "QKNj50dIZ-7g91",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "ScmJi_uFNmDfk1",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-4Jd4YjGjIL",
|
||||
"timestamp": "2026-03-26T12:55:04.731Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "gpJGGWJgrauUlp",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "UD7e9LmcNMMab6",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-7awjIpz1KA",
|
||||
"timestamp": "2026-03-26T15:06:11.446Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "l3XCZBVJI8DMb6",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "tmM6DRRQ-FM6Ef",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "xrLAKFpmR0k_fB",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,135 @@
|
||||
{
|
||||
"run_id": "eval-86xJU1J7RH",
|
||||
"timestamp": "2026-03-26T12:55:30.782Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "BCfZGlUf_7WrzX",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "hDwHzKUEvGv_WH",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "7ZV0K25MkOXdrL",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-8HiKD7tzkR",
|
||||
"timestamp": "2026-03-26T14:55:02.675Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "nAAz2ScCcfd7ML",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "fhy4j7cjoB2lTX",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-94ypeytPFD",
|
||||
"timestamp": "2026-03-26T14:48:14.207Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "JWodHRnyHMbmXF",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Pxo_T9ekuyP7k_",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-9V1JK90r5M",
|
||||
"timestamp": "2026-03-26T12:55:31.352Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "v2KqGYhtmtYcgC",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "cL_lvtjUZLKwXk",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-AKJhoo1T06",
|
||||
"timestamp": "2026-03-26T14:50:28.656Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Pz3SajCDpZUJZi",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "s8NQvLo0jtmRMg",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "SmjVXdYHtAfh78",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,135 @@
|
||||
{
|
||||
"run_id": "eval-DBE79isARo",
|
||||
"timestamp": "2026-03-26T13:16:13.683Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "xNtCuag50SXqLa",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "N7w7OEJH1AZQVj",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "Sh64q_20hIUmH0",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-ENsXL6tmGP",
|
||||
"timestamp": "2026-03-26T14:54:03.188Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "233Vomq5bQ3Etj",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "1hNbLs5GT0Zb7w",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-HK1iqqUDF3",
|
||||
"timestamp": "2026-03-26T14:55:02.664Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "-_6kti9x1EAUA0",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "X1KFs_uEDRAn70",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-I9rx4KE5gR",
|
||||
"timestamp": "2026-03-26T15:06:12.667Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "u_fv2077U9ibdi",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "X6i-Ez8mtzYyK2",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-ILx0ihm-Lw",
|
||||
"timestamp": "2026-03-26T14:26:14.249Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "gdQmHxalu1cmCy",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "GvpfTlrn6ZR1Z6",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "OUOpu8YXjUFck7",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-LYEATvWSQn",
|
||||
"timestamp": "2026-03-26T14:37:56.811Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "OGz5dN9-MYy1G8",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "UMEWQM9gAsokmI",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-NdHWJrLVPY",
|
||||
"timestamp": "2026-03-26T12:55:31.352Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "oc1AOWxwStd60g",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Aipiy0hn6HL2uY",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-O360CBYk6a",
|
||||
"timestamp": "2026-03-26T14:59:11.293Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "eIdR0sgRZ_tBSt",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "5tQorln1n1S3yD",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,135 @@
|
||||
{
|
||||
"run_id": "eval-OP8Go8cQU9",
|
||||
"timestamp": "2026-03-26T13:14:25.915Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "PO59mE6XvvZ3k-",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "8Z1GlyNzIH8r6S",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "gPRGR71mhUHYpy",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-PKpQU8_Mgt",
|
||||
"timestamp": "2026-03-26T14:48:14.219Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "-nxxWJd40iccfx",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "cpoY1EIOWYgaZQ",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-Pm5Ja_gY1X",
|
||||
"timestamp": "2026-03-26T12:55:35.823Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "7aEzzfvNPh5iIJ",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "hEQpwfmK5kkJ1R",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-Pwz7xRnbx0",
|
||||
"timestamp": "2026-03-26T12:55:35.822Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "sQv-L0-0hVtFHA",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "zmI3zT9qCn5yAi",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-QFGmXAHVxo",
|
||||
"timestamp": "2026-03-26T13:16:30.497Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "kLuH1Ts0Ndt7zi",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "uPTFOfSJRHcZEH",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-QTRrUNA92U",
|
||||
"timestamp": "2026-03-26T13:16:30.501Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "0jy-9_PxHtWE9o",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "FCWOVWFpJeo4Hg",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-QnXjMWuNG4",
|
||||
"timestamp": "2026-03-26T14:54:02.477Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "sI0z4HgYaC0hwE",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "O68FZ2L5CkIkRC",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "BcvCxZlN6oRSPt",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-T5ubfTDRvc",
|
||||
"timestamp": "2026-03-26T15:04:48.689Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "5FUopKIDNLH9x3",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "tiSeEFAbOlH4HW",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-UrYNuoIte3",
|
||||
"timestamp": "2026-03-26T15:04:46.343Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Ky7rKv9Qi2LzoJ",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "EZtK3enHi9g_LM",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "Wu0Y_-rXFh68tp",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,135 @@
|
||||
{
|
||||
"run_id": "eval-VBCi_Qxr5Y",
|
||||
"timestamp": "2026-03-26T13:16:29.901Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "7wuLnq4BH3piOu",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "73aohob638bDuA",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "WjiJJxc9QvCpvV",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-WeSd-iJ_P_",
|
||||
"timestamp": "2026-03-26T14:55:01.958Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "9kfhdQNrIURTjL",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "Ph7s_xUUKPC96z",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "hdlurNdtueZ0DX",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-X1q04xpajb",
|
||||
"timestamp": "2026-03-26T13:14:26.541Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "G86usNUuaWpz69",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "9r87IyEYbgT-dl",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-ahKw2z3DNd",
|
||||
"timestamp": "2026-03-26T13:16:14.290Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "R4oqbuO7AZVCD6",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "TDIxd1uZGRFF40",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-awkhM6kKyT",
|
||||
"timestamp": "2026-03-26T14:29:32.762Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Yv3aYwxmxBlAyJ",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "XRBd2kl7qFE8aG",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "5yW64FSU_IjWOY",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-br2XpjV0IF",
|
||||
"timestamp": "2026-03-26T14:54:03.186Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "DIyGK1CkPnPicG",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "qrabd_Br0Ccs15",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,135 @@
|
||||
{
|
||||
"run_id": "eval-coE2JZYmvV",
|
||||
"timestamp": "2026-03-26T12:55:03.368Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "XuSyxnPhZ7P_PW",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "4sdBIrJ3fn5aqX",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "9YHZELcCqUbAzQ",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-csS0kkZmz9",
|
||||
"timestamp": "2026-03-26T17:04:54.186Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "YAW8P_gnmEHbmA",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "wq0q7oxcCBNkY6",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "vbjwJRq69VeHm5",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-dtpCo4sMdc",
|
||||
"timestamp": "2026-03-26T15:06:12.670Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "wFRYwzhrxkIZ9l",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "1xHDWvJ7ZrJbB5",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-e-1UopyweC",
|
||||
"timestamp": "2026-03-26T13:16:14.288Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "5mHNGyM4epF9AB",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "unZlAEc3PGwu8l",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-fNnAM4rRXw",
|
||||
"timestamp": "2026-03-26T14:26:14.869Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "2SO3wpAWN_CpSC",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "9YxDVbJdguj2rx",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-fTyrcI8ATt",
|
||||
"timestamp": "2026-03-26T13:14:26.535Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Ynl9X1eX9phUX_",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "JYnY80AE-3YW0R",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-kZZ5rtfv4X",
|
||||
"timestamp": "2026-03-26T14:26:14.890Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "o6eQLf5SO1Gf2p",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "HKJlZl2REXlhIa",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-nCABoTHSuw",
|
||||
"timestamp": "2026-03-26T14:37:55.515Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Bcb_BEMUY56h-G",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "8hm25X41b9Fm46",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "ouqPxsmQmXYeoB",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,135 @@
|
||||
{
|
||||
"run_id": "eval-qo49IgK8zT",
|
||||
"timestamp": "2026-03-26T12:55:34.412Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "wKoH1Fcro5huCS",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "oFA0_O1Dd1EEnA",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "2HHlXRriLdK-hF",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-rgZvv7GVlh",
|
||||
"timestamp": "2026-03-26T14:50:29.357Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "DFhLxldnnEi813",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "THV16s3zq8ch9N",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-uno0zByXlr",
|
||||
"timestamp": "2026-03-26T14:59:11.292Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "tFalCYEY3jS79i",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "hEK6K6QrsUgnnu",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,136 @@
|
||||
{
|
||||
"run_id": "eval-vBRt4_zi8o",
|
||||
"timestamp": "2026-03-26T14:48:13.117Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 3
|
||||
},
|
||||
"cases_total": 3,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 33.33,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 33.33,
|
||||
"routed_fragment_rate": 33.33,
|
||||
"no_route_fragment_rate": 66.67,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 3,
|
||||
"checks_passed": 3
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 1,
|
||||
"no_route": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 1,
|
||||
"out_of_scope": 1,
|
||||
"clarification": 1
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь хвосты по поставщикам и разложи цепочку",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "jcgq3Vum-vmvtb",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Как вообще по ФСБУ",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 1,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "out_of_scope",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "out_of_scope",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "zb3PnxSTYYCKtT",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-003",
|
||||
"raw_question": "Покажи топ рисков за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": false,
|
||||
"scope_confidence": "low",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 0,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 1,
|
||||
"fallback_type": "clarification",
|
||||
"predicted_route_status": "no_route",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": "insufficient_specificity",
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 0,
|
||||
"trace_id": "07LQK4g2WMJmJM",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-vrtzSgGK5J",
|
||||
"timestamp": "2026-03-26T14:37:56.801Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "sUs-83-P-FBJC5",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "nf7HAY_H_yIRz8",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-wO0GIYPaNz",
|
||||
"timestamp": "2026-03-26T12:55:04.737Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "ozP1_JjXjrXdNb",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "gKP8y15WrKRjM6",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-wPEGrtVlnN",
|
||||
"timestamp": "2026-03-26T17:04:55.445Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "9RsDbvSEMVzwy9",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "gDKgk0Yj_UhrIX",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-wnexicBv1L",
|
||||
"timestamp": "2026-03-26T17:04:55.449Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "LChjIRCQvUVBJI",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по счету 97",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "jBg8Projzr0MZ0",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-xwO4M4uHeu",
|
||||
"timestamp": "2026-03-26T14:50:29.376Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "Eh2F-VtWbDijzY",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "4-DS8kKXsGNGSv",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
@@ -0,0 +1,111 @@
|
||||
{
|
||||
"run_id": "eval-ynin-rwW6N",
|
||||
"timestamp": "2026-03-26T14:29:33.467Z",
|
||||
"mode": "single-pass-strict",
|
||||
"use_mock": true,
|
||||
"prompt_version": "normalizer_v2_0_2",
|
||||
"schema_version": "v2_0_2",
|
||||
"dataset": {
|
||||
"source": "inline_raw_questions",
|
||||
"file": null,
|
||||
"raw_questions_count": 2
|
||||
},
|
||||
"cases_total": 2,
|
||||
"metrics": {
|
||||
"schema_validation_pass_rate": 100,
|
||||
"scope_detection_accuracy": null,
|
||||
"scope_in_scope_rate": 100,
|
||||
"multi_intent_detected_rate": 0,
|
||||
"clarification_required_rate": 0,
|
||||
"avg_fragments_per_message": 1,
|
||||
"out_of_scope_fragment_rate": 0,
|
||||
"routed_fragment_rate": 100,
|
||||
"no_route_fragment_rate": 0,
|
||||
"route_resolution_accuracy": null,
|
||||
"no_route_precision": null,
|
||||
"false_no_route_rate": null,
|
||||
"execution_state_consistency_rate": 100,
|
||||
"executable_with_soft_assumptions_rate": 100,
|
||||
"soft_assumption_used_fragment_rate": 100,
|
||||
"clarification_precision": null,
|
||||
"clarification_recall": null,
|
||||
"false_clarification_rate": null
|
||||
},
|
||||
"budget": {
|
||||
"requests_total": 0,
|
||||
"retries_used": 0
|
||||
},
|
||||
"clarification_eval": {
|
||||
"labeled_cases": 0,
|
||||
"true_positive": 0,
|
||||
"false_positive": 0,
|
||||
"false_negative": 0
|
||||
},
|
||||
"route_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0,
|
||||
"expected_routed_cases": 0,
|
||||
"no_route_true_positive": 0,
|
||||
"no_route_false_positive": 0
|
||||
},
|
||||
"scope_eval": {
|
||||
"labeled_cases": 0,
|
||||
"correct_cases": 0
|
||||
},
|
||||
"execution_state_eval": {
|
||||
"checks_total": 2,
|
||||
"checks_passed": 2
|
||||
},
|
||||
"route_distribution": {
|
||||
"store_feature_risk": 2
|
||||
},
|
||||
"fallback_distribution": {
|
||||
"none": 2
|
||||
},
|
||||
"results": [
|
||||
{
|
||||
"case_id": "BQ-001",
|
||||
"raw_question": "Проверь счет 60 за июнь 2020",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "491D4elziSwn2k",
|
||||
"request_count_for_case": 0
|
||||
},
|
||||
{
|
||||
"case_id": "BQ-002",
|
||||
"raw_question": "Покажи риски по НДС и по закрытию",
|
||||
"validation_passed": true,
|
||||
"message_in_scope": true,
|
||||
"scope_confidence": "high",
|
||||
"contains_multiple_tasks": false,
|
||||
"fragments_total": 1,
|
||||
"in_scope_fragments": 1,
|
||||
"out_of_scope_fragments": 0,
|
||||
"unclear_fragments": 0,
|
||||
"fallback_type": "none",
|
||||
"predicted_route_status": "routed",
|
||||
"expected_route_status": null,
|
||||
"predicted_no_route_reason": null,
|
||||
"expected_no_route_reason": null,
|
||||
"predicted_clarification_required": false,
|
||||
"expected_clarification_required": null,
|
||||
"executable_with_soft_assumptions_fragments": 1,
|
||||
"trace_id": "_iYyGaR7yizo7k",
|
||||
"request_count_for_case": 0
|
||||
}
|
||||
]
|
||||
}
|
||||
Reference in New Issue
Block a user