Закрепить агентный replay по tUy7 smalltalk и НДС

This commit is contained in:
dctouch 2026-06-01 20:59:42 +03:00
parent 5426f9a24c
commit 599ca7d055
4 changed files with 504 additions and 0 deletions

View File

@ -0,0 +1,172 @@
{
"schema_version": "domain_truth_harness_spec_v1",
"scenario_id": "agent_tuy7_smalltalk_vat_purchase_repair",
"domain": "assistant_stage1_agent_repair",
"title": "AGENT repair replay for tUy7 smalltalk and VAT purchase-date anchor",
"description": "Targeted replay for assistant-stage1-tUy7-NlyI5 defects: first-turn smalltalk must stay deterministic without backend_error, and VAT follow-up over the selected inventory item must keep the purchase-date anchor instead of answering with a generic organization tax period.",
"bindings": {},
"steps": [
{
"step_id": "s01_smalltalk",
"title": "First turn smalltalk does not hit backend error",
"question": "приветик - че как там дела",
"required_answer_patterns_any": [
"(?i)привет|дела|помоч|готов"
],
"forbidden_answer_patterns": [
"(?i)внутренн.*ошиб",
"(?i)контуров LLM",
"(?i)OpenAI API key",
"(?i)backend_error"
],
"criticality": "critical",
"semantic_tags": [
"smalltalk_entry",
"backend_error_guard"
]
},
{
"step_id": "s02_capability",
"title": "Capability meta remains useful after smalltalk",
"question": "расскажи что можешь интересного",
"required_answer_patterns_all": [
"(?i)могу|помог",
"(?i)ндс|склад|контрагент|долг|документ"
],
"forbidden_answer_patterns": [
"(?i)backend_error",
"(?i)capability_id",
"(?i)runtime_"
],
"criticality": "important",
"semantic_tags": [
"capability_meta"
]
},
{
"step_id": "s03_select_org",
"title": "Organization selection keeps working scope",
"question": "АЛЬТЕРНАТИВА",
"required_answer_patterns_any": [
"(?i)Альтернатива Плюс",
"(?i)фиксир|рабоч.*организац"
],
"forbidden_answer_patterns": [
"(?i)backend_error"
],
"criticality": "important",
"semantic_tags": [
"organization_selection"
]
},
{
"step_id": "s04_stock_current",
"title": "Inventory root establishes stock context after organization selection",
"question": "кайф - что там на складе по остаткам?",
"required_answer_patterns_all": [
"(?i)склад",
"(?i)остат|позици"
],
"forbidden_answer_patterns": [
"(?i)backend_error",
"(?i)capability_id"
],
"criticality": "important",
"semantic_tags": [
"inventory",
"context_root"
]
},
{
"step_id": "s05_history_capability",
"title": "Historical inventory capability follow-up",
"question": "а исторические остатки на другие даты умеешь?",
"required_answer_patterns_all": [
"(?i)историческ|дат|период",
"(?i)остат"
],
"forbidden_answer_patterns": [
"(?i)backend_error"
],
"criticality": "important",
"semantic_tags": [
"inventory_history_capability"
]
},
{
"step_id": "s06_july_2017",
"title": "Historical inventory July 2017",
"question": "давай на июль 2017",
"required_answer_patterns_all": [
"(?i)31\\.07\\.2017|июл",
"(?i)Рабочая станция"
],
"forbidden_answer_patterns": [
"(?i)backend_error"
],
"criticality": "important",
"semantic_tags": [
"inventory_history"
]
},
{
"step_id": "s07_march_2016",
"title": "Historical inventory March 2016",
"question": "март 2016",
"required_answer_patterns_all": [
"(?i)31\\.03\\.2016|март",
"(?i)Рабочая станция"
],
"forbidden_answer_patterns": [
"(?i)backend_error"
],
"criticality": "important",
"semantic_tags": [
"inventory_history",
"temporal_carryover"
]
},
{
"step_id": "s08_selected_purchase_trace",
"title": "Selected workstation purchase trace establishes purchase-date bundle",
"question": "По выбранному объекту \"Рабочая станция универсального специалиста (индивидуальное изготовление)\": где взяли это?",
"required_answer_patterns_all": [
"(?i)Рабочая станция",
"(?i)поставщик|поступлен|закуп",
"(?i)2015|05\\.02\\.2015|22\\.07\\.2015"
],
"forbidden_answer_patterns": [
"(?i)backend_error",
"(?i)capability_id"
],
"criticality": "critical",
"semantic_tags": [
"selected_object",
"purchase_trace"
]
},
{
"step_id": "s09_vat_purchase_date",
"title": "VAT follow-up keeps selected item purchase-date anchor",
"question": "ндс можешь прикинуть на дату покупки рабочей станции?",
"required_answer_patterns_all": [
"(?i)НДС",
"(?i)покуп|закуп|поступлен",
"(?i)2015|05\\.02\\.2015|22\\.07\\.2015"
],
"forbidden_answer_patterns": [
"(?i)^Коротко:",
"(?i)2\\.631\\.388,00",
"(?i)общ.*налогов.*позици",
"(?i)backend_error",
"(?i)capability_id"
],
"criticality": "critical",
"semantic_tags": [
"vat",
"purchase_date_anchor",
"selected_object_followup"
]
}
]
}

View File

@ -1,4 +1,61 @@
[ [
{
"generation_id": "gen-ag06011759-6d66c2",
"created_at": "2026-06-01T17:59:07+00:00",
"mode": "saved_user_sessions",
"title": "AGENT repair replay for tUy7 smalltalk and VAT purchase-date anchor",
"count": 9,
"domain": "assistant_stage1_agent_repair",
"questions": [
"приветик - че как там дела",
"расскажи что можешь интересного",
"АЛЬТЕРНАТИВА",
"кайф - что там на складе по остаткам?",
"а исторические остатки на другие даты умеешь?",
"давай на июль 2017",
"март 2016",
"По выбранному объекту \"Рабочая станция универсального специалиста (индивидуальное изготовление)\": где взяли это?",
"ндс можешь прикинуть на дату покупки рабочей станции?"
],
"generated_by": "codex_agent",
"saved_case_set_file": "assistant_autogen_saved_user_sessions_20260601175907_gen-ag06011759-6d66c2.json",
"context": {
"llm_provider": null,
"model": null,
"assistant_prompt_version": null,
"decomposition_prompt_version": null,
"prompt_fingerprint": null,
"autogen_personality_id": null,
"autogen_personality_prompt": null,
"source_session_id": null,
"saved_session_file": "assistant_saved_session_20260601175907_gen-ag06011759-6d66c2.json",
"saved_case_set_kind": "agent_semantic_scenario",
"agent_run": true,
"agent_focus": "Targeted replay for assistant-stage1-tUy7-NlyI5 defects: first-turn smalltalk must stay deterministic without backend_error, and VAT follow-up over the selected inventory item must keep the purchase-date anchor instead of answering with a generic organization tax period.",
"architecture_phase": "turnaround_11",
"source_spec_file": "X:\\1C\\NDC_1C\\docs\\orchestration\\agent_tuy7_smalltalk_vat_purchase_repair.json",
"scenario_id": "agent_tuy7_smalltalk_vat_purchase_repair",
"semantic_tags": [
"backend_error_guard",
"capability_meta",
"context_root",
"inventory",
"inventory_history",
"inventory_history_capability",
"organization_selection",
"purchase_date_anchor",
"purchase_trace",
"selected_object",
"selected_object_followup",
"smalltalk_entry",
"temporal_carryover",
"vat"
],
"validation_status": "accepted_live_replay",
"validated_run_dir": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2",
"saved_after_validated_replay": true
}
},
{ {
"generation_id": "gen-ag06011720-49fcf4", "generation_id": "gen-ag06011720-49fcf4",
"created_at": "2026-06-01T17:20:01+00:00", "created_at": "2026-06-01T17:20:01+00:00",

View File

@ -0,0 +1,223 @@
{
"saved_at": "2026-06-01T17:59:07+00:00",
"generation_id": "gen-ag06011759-6d66c2",
"mode": "saved_user_sessions",
"title": "AGENT repair replay for tUy7 smalltalk and VAT purchase-date anchor",
"agent_run": true,
"questions": [
"приветик - че как там дела",
"расскажи что можешь интересного",
"АЛЬТЕРНАТИВА",
"кайф - что там на складе по остаткам?",
"а исторические остатки на другие даты умеешь?",
"давай на июль 2017",
"март 2016",
"По выбранному объекту \"Рабочая станция универсального специалиста (индивидуальное изготовление)\": где взяли это?",
"ндс можешь прикинуть на дату покупки рабочей станции?"
],
"metadata": {
"assistant_prompt_version": null,
"decomposition_prompt_version": null,
"prompt_fingerprint": null,
"agent_focus": "Targeted replay for assistant-stage1-tUy7-NlyI5 defects: first-turn smalltalk must stay deterministic without backend_error, and VAT follow-up over the selected inventory item must keep the purchase-date anchor instead of answering with a generic organization tax period.",
"architecture_phase": "turnaround_11",
"source_spec_file": "X:\\1C\\NDC_1C\\docs\\orchestration\\agent_tuy7_smalltalk_vat_purchase_repair.json",
"scenario_id": "agent_tuy7_smalltalk_vat_purchase_repair",
"semantic_tags": [
"backend_error_guard",
"capability_meta",
"context_root",
"inventory",
"inventory_history",
"inventory_history_capability",
"organization_selection",
"purchase_date_anchor",
"purchase_trace",
"selected_object",
"selected_object_followup",
"smalltalk_entry",
"temporal_carryover",
"vat"
],
"validation_status": "accepted_live_replay",
"validated_run_dir": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2",
"saved_after_validated_replay": true,
"save_gate": {
"schema_version": "agent_semantic_save_gate_v1",
"validation_status": "accepted_live_replay",
"validated_run_dir": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2",
"final_status": "accepted",
"review_overall_status": "pass",
"business_overall_status": "pass",
"steps_total": 9,
"steps_passed": 9,
"steps_failed": 0,
"steps_with_business_failures": 0,
"steps_with_business_warnings": 0,
"acceptance_gate_passed": true,
"effective_runtime": {
"manifest_path": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2\\effective_runtime.json",
"runner": "domain_truth_harness.run-live",
"git_sha": "5426f9a24ccea10813fa4f3e63c80a6db59702b1",
"backend_url": "http://127.0.0.1:8787",
"mcp_proxy_url": "http://127.0.0.1:6003",
"llm_provider": "local",
"llm_model": "unsloth/qwen3-30b-a3b-instruct-2507",
"temperature": 0.8,
"max_output_tokens": 900,
"prompt_version": "normalizer_v2_0_2",
"prompt_source": "file",
"prompt_hash": "f36e36a5e491cd24511b380e0c0059cb01aea5e2af738fb9f01671cd291735c1",
"prompt_registry_status": "pass"
},
"saved_after_validated_replay": true
}
},
"source_session_id": null,
"session": {
"session_id": null,
"mode": "agent_semantic_run",
"items": [
{
"message_id": "agent-user-001",
"role": "user",
"text": "приветик - че как там дела",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-002",
"role": "user",
"text": "расскажи что можешь интересного",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-003",
"role": "user",
"text": "АЛЬТЕРНАТИВА",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-004",
"role": "user",
"text": "кайф - что там на складе по остаткам?",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-005",
"role": "user",
"text": "а исторические остатки на другие даты умеешь?",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-006",
"role": "user",
"text": "давай на июль 2017",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-007",
"role": "user",
"text": "март 2016",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-008",
"role": "user",
"text": "По выбранному объекту \"Рабочая станция универсального специалиста (индивидуальное изготовление)\": где взяли это?",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
},
{
"message_id": "agent-user-009",
"role": "user",
"text": "ндс можешь прикинуть на дату покупки рабочей станции?",
"created_at": "2026-06-01T17:59:07+00:00",
"reply_type": null,
"trace_id": null,
"debug": null
}
],
"agent_run": true,
"metadata": {
"assistant_prompt_version": null,
"decomposition_prompt_version": null,
"prompt_fingerprint": null,
"agent_focus": "Targeted replay for assistant-stage1-tUy7-NlyI5 defects: first-turn smalltalk must stay deterministic without backend_error, and VAT follow-up over the selected inventory item must keep the purchase-date anchor instead of answering with a generic organization tax period.",
"architecture_phase": "turnaround_11",
"source_spec_file": "X:\\1C\\NDC_1C\\docs\\orchestration\\agent_tuy7_smalltalk_vat_purchase_repair.json",
"scenario_id": "agent_tuy7_smalltalk_vat_purchase_repair",
"semantic_tags": [
"backend_error_guard",
"capability_meta",
"context_root",
"inventory",
"inventory_history",
"inventory_history_capability",
"organization_selection",
"purchase_date_anchor",
"purchase_trace",
"selected_object",
"selected_object_followup",
"smalltalk_entry",
"temporal_carryover",
"vat"
],
"validation_status": "accepted_live_replay",
"validated_run_dir": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2",
"saved_after_validated_replay": true,
"save_gate": {
"schema_version": "agent_semantic_save_gate_v1",
"validation_status": "accepted_live_replay",
"validated_run_dir": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2",
"final_status": "accepted",
"review_overall_status": "pass",
"business_overall_status": "pass",
"steps_total": 9,
"steps_passed": 9,
"steps_failed": 0,
"steps_with_business_failures": 0,
"steps_with_business_warnings": 0,
"acceptance_gate_passed": true,
"effective_runtime": {
"manifest_path": "artifacts\\domain_runs\\agent_tuy7_smalltalk_vat_purchase_repair_p2\\effective_runtime.json",
"runner": "domain_truth_harness.run-live",
"git_sha": "5426f9a24ccea10813fa4f3e63c80a6db59702b1",
"backend_url": "http://127.0.0.1:8787",
"mcp_proxy_url": "http://127.0.0.1:6003",
"llm_provider": "local",
"llm_model": "unsloth/qwen3-30b-a3b-instruct-2507",
"temperature": 0.8,
"max_output_tokens": 900,
"prompt_version": "normalizer_v2_0_2",
"prompt_source": "file",
"prompt_hash": "f36e36a5e491cd24511b380e0c0059cb01aea5e2af738fb9f01671cd291735c1",
"prompt_registry_status": "pass"
},
"saved_after_validated_replay": true
}
}
}
}

View File

@ -0,0 +1,52 @@
{
"suite_id": "assistant_saved_session_gen-ag06011759-6d66c2",
"suite_version": "0.1.0",
"schema_version": "assistant_saved_session_suite_v0_1",
"generated_at": "2026-06-01T17:59:07+00:00",
"generation_id": "gen-ag06011759-6d66c2",
"mode": "saved_user_sessions",
"title": "AGENT repair replay for tUy7 smalltalk and VAT purchase-date anchor",
"domain": "assistant_stage1_agent_repair",
"scenario_count": 1,
"case_ids": [
"SAVED-001"
],
"cases": [
{
"case_id": "SAVED-001",
"scenario_tag": "agent_saved_user_sessions",
"title": "AGENT repair replay for tUy7 smalltalk and VAT purchase-date anchor",
"question_type": "followup",
"broadness_level": "medium",
"turns": [
{
"user_message": "приветик - че как там дела"
},
{
"user_message": "расскажи что можешь интересного"
},
{
"user_message": "АЛЬТЕРНАТИВА"
},
{
"user_message": "кайф - что там на складе по остаткам?"
},
{
"user_message": "а исторические остатки на другие даты умеешь?"
},
{
"user_message": "давай на июль 2017"
},
{
"user_message": "март 2016"
},
{
"user_message": "По выбранному объекту \"Рабочая станция универсального специалиста (индивидуальное изготовление)\": где взяли это?"
},
{
"user_message": "ндс можешь прикинуть на дату покупки рабочей станции?"
}
]
}
]
}