АДРЕСНЫЙ РЕЖИМ - авторан история - юи + адресный рендер прогонов в реалтайме

llm_normalizer/frontend/src/api/client.ts

@@ -1,4 +1,7 @@
 import type {
+  AsyncEvalRunStartResponse,
+  AsyncEvalRunStatusResponse,
+  AutoGenPersonalityCatalogResponse,
   AutoGenHistoryResponse,
   AutoGenMode,
   AutoRunAnnotationsResponse,
@@ -155,6 +158,8 @@ export const apiClient = {
     mode?: "standard" | "single-pass-strict";
     caseSetFile?: string;
     rawQuestions?: string;
+    evalTarget?: "normalizer" | "assistant_stage1" | "assistant_stage2" | "assistant_p0";
+    compareWithReportFile?: string;
   }): Promise<{ ok: boolean; report: unknown }> {
     return request("/eval/run", {
       method: "POST",
@@ -176,11 +181,58 @@ export const apiClient = {
         useMock: Boolean(input.useMock),
         mode: input.mode ?? "standard",
         caseSetFile: input.caseSetFile,
-        rawQuestions: input.rawQuestions
+        rawQuestions: input.rawQuestions,
+        eval_target: input.evalTarget,
+        compare_with_report_file: input.compareWithReportFile
       })
     });
   },
 
+  async startEvalRunAsync(input: {
+    connection: ConnectionState;
+    prompts: PromptState;
+    promptVersion?: string;
+    caseIds?: string[];
+    useMock?: boolean;
+    mode?: "standard" | "single-pass-strict";
+    caseSetFile?: string;
+    rawQuestions?: string;
+    evalTarget?: "normalizer" | "assistant_stage1" | "assistant_stage2" | "assistant_p0";
+    compareWithReportFile?: string;
+    questions?: string[];
+  }): Promise<AsyncEvalRunStartResponse> {
+    return request("/eval/run-async/start", {
+      method: "POST",
+      body: JSON.stringify({
+        normalizeConfig: {
+          llmProvider: input.connection.llmProvider,
+          apiKey: input.connection.apiKey,
+          model: input.connection.model,
+          baseUrl: input.connection.baseUrl,
+          temperature: input.connection.temperature,
+          maxOutputTokens: input.connection.maxOutputTokens,
+          promptVersion: input.promptVersion,
+          systemPrompt: input.prompts.systemPrompt,
+          developerPrompt: input.prompts.developerPrompt,
+          domainPrompt: input.prompts.domainPrompt,
+          fewShotExamples: input.prompts.fewShotExamples
+        },
+        caseIds: input.caseIds,
+        useMock: Boolean(input.useMock),
+        mode: input.mode ?? "standard",
+        caseSetFile: input.caseSetFile,
+        rawQuestions: input.rawQuestions,
+        eval_target: input.evalTarget,
+        compare_with_report_file: input.compareWithReportFile,
+        questions: input.questions
+      })
+    });
+  },
+
+  async loadEvalRunAsyncStatus(jobId: string): Promise<AsyncEvalRunStatusResponse> {
+    return request(`/eval/run-async/${encodeURIComponent(jobId)}`);
+  },
+
   async startRun(): Promise<{ ok: boolean; run: RuntimeRun; runId: string; sessionId: string; status: string }> {
     return request("/accounting-agent/v1/runs/start", {
       method: "POST",
@@ -321,6 +373,20 @@ export const apiClient = {
     });
   },
 
+  async updateAutoRunAnnotation(input: {
+    annotation_id: string;
+    resolved: boolean;
+    resolved_by?: string;
+  }): Promise<{ ok: boolean; annotation: AutoRunAnnotationRecord; case_annotation_stats: { count: number; latest_at: string | null; avg_rating: number | null } | null }> {
+    return request(`/autoruns/annotations/${encodeURIComponent(input.annotation_id)}`, {
+      method: "PATCH",
+      body: JSON.stringify({
+        resolved: input.resolved,
+        resolved_by: input.resolved_by
+      })
+    });
+  },
+
   async loadAutoRunPostAnalysis(input?: {
     run_id?: string;
     limit_per_queue?: number;
@@ -359,12 +425,24 @@ export const apiClient = {
     return request(`/autoruns/autogen/history${query ? `?${query}` : ""}`);
   },
 
+  async loadAutoRunAutogenPersonalityCatalog(): Promise<AutoGenPersonalityCatalogResponse> {
+    return request("/autoruns/autogen/personality-catalog");
+  },
+
   async generateAutoRunQuestions(input: {
     mode: AutoGenMode;
     count: number;
     domain?: string;
     persist_to_eval_cases?: boolean;
     generated_by?: string;
+    llm?: {
+      llm_provider?: "openai" | "local";
+      api_key?: string;
+      model?: string;
+      base_url?: string;
+      temperature?: number;
+      max_output_tokens?: number;
+    };
     context?: {
       llm_provider?: string;
       model?: string;

llm_normalizer/frontend/src/components/ConnectionPanel.tsx

@@ -1,3 +1,4 @@
+import { useEffect, useState } from "react";
 import { PanelFrame } from "./PanelFrame";
 import type { ConnectionState } from "../state/types";
 
@@ -26,6 +27,46 @@ export function ConnectionPanel({
 }: ConnectionPanelProps) {
   const isLocal = value.llmProvider === "local";
   const modelInCatalog = modelOptions.includes(value.model);
+  const [temperatureInput, setTemperatureInput] = useState(String(value.temperature));
+  const [maxOutputTokensInput, setMaxOutputTokensInput] = useState(String(value.maxOutputTokens));
+
+  useEffect(() => {
+    setTemperatureInput(String(value.temperature));
+  }, [value.temperature]);
+
+  useEffect(() => {
+    setMaxOutputTokensInput(String(value.maxOutputTokens));
+  }, [value.maxOutputTokens]);
+
+  const commitTemperatureInput = (raw: string) => {
+    const normalized = raw.replace(",", ".").trim();
+    if (!normalized) {
+      setTemperatureInput(String(value.temperature));
+      return;
+    }
+    const parsed = Number(normalized);
+    if (!Number.isFinite(parsed)) {
+      setTemperatureInput(String(value.temperature));
+      return;
+    }
+    onChange({ ...value, temperature: parsed });
+    setTemperatureInput(String(parsed));
+  };
+
+  const commitMaxOutputTokensInput = (raw: string) => {
+    const normalized = raw.trim();
+    if (!normalized) {
+      setMaxOutputTokensInput(String(value.maxOutputTokens));
+      return;
+    }
+    const parsed = Number.parseInt(normalized, 10);
+    if (!Number.isFinite(parsed) || parsed <= 0) {
+      setMaxOutputTokensInput(String(value.maxOutputTokens));
+      return;
+    }
+    onChange({ ...value, maxOutputTokens: parsed });
+    setMaxOutputTokensInput(String(parsed));
+  };
 
   return (
     <PanelFrame
@@ -108,8 +149,14 @@ export function ConnectionPanel({
           <input
             type="number"
             step="0.1"
-            value={value.temperature}
-            onChange={(event) => onChange({ ...value, temperature: Number(event.target.value) })}
+            value={temperatureInput}
+            onChange={(event) => setTemperatureInput(event.target.value)}
+            onBlur={(event) => commitTemperatureInput(event.target.value)}
+            onKeyDown={(event) => {
+              if (event.key === "Enter") {
+                commitTemperatureInput((event.target as HTMLInputElement).value);
+              }
+            }}
           />
         </label>
 
@@ -117,8 +164,14 @@ export function ConnectionPanel({
           Max output tokens
           <input
             type="number"
-            value={value.maxOutputTokens}
-            onChange={(event) => onChange({ ...value, maxOutputTokens: Number(event.target.value) })}
+            value={maxOutputTokensInput}
+            onChange={(event) => setMaxOutputTokensInput(event.target.value)}
+            onBlur={(event) => commitMaxOutputTokensInput(event.target.value)}
+            onKeyDown={(event) => {
+              if (event.key === "Enter") {
+                commitMaxOutputTokensInput((event.target as HTMLInputElement).value);
+              }
+            }}
           />
         </label>
       </div>

llm_normalizer/frontend/src/state/types.ts

@@ -188,6 +188,9 @@ export interface AutoRunAnnotationRecord {
   comment: string;
   manual_case_decision: ManualCaseDecision;
   annotation_author: string | null;
+  resolved: boolean;
+  resolved_at: string | null;
+  resolved_by: string | null;
   created_at: string;
   updated_at: string;
   context: {
@@ -198,6 +201,8 @@ export interface AutoRunAnnotationRecord {
     prompt_version: string | null;
     domain: string | null;
     query_class: string | null;
+    question_text: string | null;
+    answer_text: string | null;
   };
 }
 
@@ -243,6 +248,8 @@ export interface AutoRunDialogMessage {
   trace_id: string | null;
   reply_type: string | null;
   message_index: number;
+  case_id?: string | null;
+  case_message_index?: number | null;
   commented: boolean;
   annotation: AutoRunAnnotationRecord | null;
 }
@@ -251,7 +258,7 @@ export interface AutoRunDialogResponse {
   ok: boolean;
   run_id: string;
   case_id: string;
-  source: "assistant_session" | "report_fallback" | "none";
+  source: "assistant_session" | "report_fallback" | "run_aggregate" | "none";
   session_id: string;
   messages: AutoRunDialogMessage[];
   decomposition: string[];
@@ -320,6 +327,8 @@ export interface AutoGenHistoryRecord {
     assistant_prompt_version: string | null;
     decomposition_prompt_version: string | null;
     prompt_fingerprint: string | null;
+    autogen_personality_id: string | null;
+    autogen_personality_prompt: string | null;
   } | null;
 }
 
@@ -329,6 +338,57 @@ export interface AutoGenHistoryResponse {
   items: AutoGenHistoryRecord[];
 }
 
+export interface AutoGenPersonalityDefinition {
+  id: string;
+  label: string;
+  domain: string | null;
+  default_prompt: string;
+  source: "built_in" | "capabilities_registry";
+}
+
+export interface AutoGenPersonalityCatalogResponse {
+  ok: boolean;
+  generated_at: string;
+  items: AutoGenPersonalityDefinition[];
+}
+
+export interface AsyncEvalRunCase {
+  case_id: string;
+  turns_total: number;
+  status: "queued" | "running" | "completed" | "failed";
+  messages: AutoRunDialogMessage[];
+}
+
+export interface AsyncEvalRunJob {
+  job_id: string;
+  status: "queued" | "running" | "completed" | "failed";
+  created_at: string;
+  updated_at: string;
+  eval_target: AutoRunTarget;
+  run_id: string;
+  case_set_file: string | null;
+  total_cases: number;
+  completed_cases: number;
+  error: string | null;
+  cases: AsyncEvalRunCase[];
+  report_summary: {
+    run_id: string | null;
+    run_timestamp: string | null;
+    score_index: number | null;
+    cases_total: number | null;
+  } | null;
+}
+
+export interface AsyncEvalRunStartResponse {
+  ok: boolean;
+  job: AsyncEvalRunJob;
+}
+
+export interface AsyncEvalRunStatusResponse {
+  ok: boolean;
+  job: AsyncEvalRunJob;
+}
+
 export type AssistantFallbackType = "none" | "out_of_scope" | "clarification" | "partial" | "unknown";
 export type AssistantReplyType =
   | "factual"

llm_normalizer/frontend/src/styles.css

@@ -741,10 +741,17 @@ button:disabled {
   padding: 10px;
   display: grid;
   gap: 5px;
+  transition: background-color 0.2s ease, box-shadow 0.2s ease;
 }
 
 .autoruns-run-item.selected {
-  border-color: var(--line-strong);
+  background: rgb(var(--rgb-active));
+  color: rgb(var(--rgb-active-text));
+  box-shadow: none;
+}
+
+.autoruns-run-item.selected .autoruns-run-meta {
+  color: rgba(var(--rgb-active-text), 0.95);
 }
 
 .autoruns-run-head,
@@ -790,7 +797,9 @@ button:disabled {
 }
 
 .autoruns-case-item.selected {
-  border-color: var(--line-strong);
+  background: rgb(var(--rgb-active));
+  color: rgb(var(--rgb-active-text));
+  box-shadow: none;
 }
 
 .autoruns-dialog-view {
@@ -831,6 +840,17 @@ button:disabled {
   gap: 8px;
 }
 
+.autoruns-msg-case-tag {
+  display: inline-flex;
+  align-items: center;
+  border-radius: 999px;
+  padding: 2px 8px;
+  font-size: 0.7rem;
+  line-height: 1;
+  color: rgb(var(--rgb-active-text));
+  background: rgba(var(--rgb-active), 0.24);
+}
+
 .autoruns-msg p {
   margin: 0;
   white-space: pre-wrap;
@@ -840,30 +860,106 @@ button:disabled {
 
 .autoruns-comment-icon {
   border: none;
-  background: rgb(var(--rgb-surface-horizontal));
-  color: var(--text-main);
-  border-radius: 999px;
-  min-width: 28px;
-  min-height: 28px;
-  padding: 0 8px;
+  background: transparent;
+  color: rgb(var(--rgb-text-main));
+  border-radius: 0;
+  min-width: 20px;
+  min-height: 20px;
+  width: 20px;
+  height: 20px;
+  padding: 0;
   line-height: 1;
   box-shadow: none;
   transform: none;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
 }
 
 .autoruns-comment-icon:hover {
-  border-color: var(--line-strong);
+  background: transparent;
+  color: rgb(var(--rgb-active));
   box-shadow: none;
   transform: none;
 }
 
 .autoruns-comment-icon.commented {
-  border-color: var(--line-strong);
-  color: var(--lime-main);
+  color: rgb(var(--rgb-active-text));
+  background: transparent;
+  box-shadow: none;
+}
+
+.comment-icon-svg {
+  width: 20px;
+  height: 20px;
+  stroke: currentColor;
+  stroke-width: 1.75;
+  stroke-linecap: round;
+  stroke-linejoin: round;
+  fill: none;
+}
+
+.comment-icon-svg .comment-icon-dot {
+  fill: currentColor;
+}
+
+.comment-icon-svg.commented {
+  fill: rgb(var(--rgb-active));
+  stroke: rgb(var(--rgb-active));
+}
+
+.autoruns-comment-icon.commented .comment-icon-dot {
+  fill: rgb(var(--rgb-active-text));
+}
+
+.autoruns-resolve-toggle {
+  border: none;
+  background: rgb(var(--rgb-surface-focus));
+  color: rgb(var(--rgb-text-main));
+  border-radius: 999px;
+  width: 22px;
+  height: 22px;
+  min-width: 22px;
+  min-height: 22px;
+  padding: 0;
+  display: inline-flex;
+  align-items: center;
+  justify-content: center;
+  box-shadow: none;
+  transform: none;
+}
+
+.autoruns-resolve-toggle:hover {
+  background: rgb(var(--rgb-surface-focus));
+  color: rgb(var(--rgb-active));
+  box-shadow: none;
+  transform: none;
+}
+
+.autoruns-resolve-toggle:disabled {
+  opacity: 0.55;
+  cursor: wait;
+}
+
+.autoruns-resolve-toggle.resolved {
   background: rgb(var(--rgb-active));
   color: rgb(var(--rgb-active-text));
 }
 
+.resolve-icon-svg {
+  width: 14px;
+  height: 14px;
+  fill: none;
+  stroke: currentColor;
+  stroke-width: 1.8;
+  stroke-linecap: round;
+  stroke-linejoin: round;
+}
+
+.resolve-icon-svg.resolved {
+  fill: currentColor;
+}
+
 .autoruns-msg-annotation {
   display: grid;
   gap: 4px;
@@ -882,6 +978,7 @@ button:disabled {
   flex: 1 1 auto;
   overflow: auto;
   padding-right: 2px;
+  margin-top: 6px;
 }
 
 .autoruns-autogen-list {
@@ -900,6 +997,17 @@ button:disabled {
   padding: 8px;
   display: grid;
   gap: 5px;
+  cursor: pointer;
+}
+
+.autoruns-autogen-item.selected {
+  background: rgb(var(--rgb-active));
+  color: rgb(var(--rgb-active-text));
+}
+
+.autoruns-autogen-item.selected .autoruns-run-meta,
+.autoruns-autogen-item.selected p {
+  color: rgba(var(--rgb-active-text), 0.95);
 }
 
 .autoruns-autogen-item header {
@@ -916,6 +1024,58 @@ button:disabled {
   font-size: 0.8rem;
 }
 
+.autoruns-generated-questions {
+  border: none;
+  border-radius: 10px;
+  background: rgb(var(--rgb-surface-horizontal));
+  padding: 8px;
+  display: grid;
+  gap: 8px;
+}
+
+.autoruns-generated-questions-head {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 8px;
+}
+
+.autoruns-generated-questions-list {
+  display: grid;
+  gap: 6px;
+  max-height: 220px;
+  overflow: auto;
+  padding-right: 2px;
+}
+
+.autoruns-generated-question-item {
+  display: flex;
+  align-items: flex-start;
+  justify-content: space-between;
+  gap: 8px;
+  border: none;
+  border-radius: 9px;
+  background: rgb(var(--rgb-surface-focus));
+  padding: 6px 8px;
+  font-size: 0.78rem;
+}
+
+.autoruns-generated-question-item span {
+  white-space: pre-wrap;
+}
+
+.autoruns-remove-question-btn {
+  flex: 0 0 auto;
+  border: none;
+  border-radius: 7px;
+  background: rgb(var(--rgb-surface-horizontal));
+  color: var(--text-main);
+  min-width: 24px;
+  height: 24px;
+  font-size: 0.75rem;
+  line-height: 1;
+}
+
 .autoruns-comment-item {
   width: 100%;
   text-align: left;
@@ -926,6 +1086,7 @@ button:disabled {
   padding: 9px;
   display: grid;
   gap: 6px;
+  cursor: pointer;
 }
 
 .autoruns-comment-item p {
@@ -936,16 +1097,48 @@ button:disabled {
 }
 
 .autoruns-comment-item.selected {
-  border-color: var(--line-strong);
+  background: rgb(var(--rgb-active));
+  color: rgb(var(--rgb-active-text));
+}
+
+.autoruns-comment-item.selected p,
+.autoruns-comment-item.selected .autoruns-run-meta,
+.autoruns-comment-item.selected .muted {
+  color: rgba(var(--rgb-active-text), 0.94);
+}
+
+.autoruns-comment-item.selected .autoruns-resolve-toggle {
+  background: rgba(var(--rgb-active-text), 0.18);
+  color: rgb(var(--rgb-active-text));
 }
 
 .autoruns-comment-head {
   display: flex;
   justify-content: space-between;
+  align-items: center;
   gap: 8px;
   font-size: 0.75rem;
 }
 
+.autoruns-comment-head-actions {
+  display: inline-flex;
+  align-items: center;
+  gap: 8px;
+}
+
+.autoruns-comment-filter-row {
+  display: grid;
+  grid-template-columns: minmax(0, 1fr) auto;
+  align-items: end;
+  gap: 10px;
+  margin-bottom: 8px;
+}
+
+.autoruns-resolved-filter-toggle {
+  min-height: 38px;
+  white-space: nowrap;
+}
+
 .autoruns-msg.assistant {
   margin-right: 12%;
 }
@@ -1096,7 +1289,8 @@ button:disabled {
 
   .autoruns-form-grid,
   .autoruns-dialog-toolbar,
-  .autoruns-stats-grid {
+  .autoruns-stats-grid,
+  .autoruns-comment-filter-row {
     grid-template-columns: 1fr;
   }
 }