feat: 增强 auto mode 的易用性 (#312)

* feat: poor 模式降级 yolo 审阅模型

* feat: 为多模块添加 Langfuse tracing 支持

在 web search、agent creation、away summary、token estimation、
skill improvement 等模块中集成 Langfuse trace，并透传至
compact/apiQueryHook/execPromptHook 等调用链。

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: 让 auto mode 记录回主 trace

* fix: reopen auto mode prompt when classifier is unavailable

* fix: 修复 auto mode 情况下, llm 报错导致弹窗也不打开的问题

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

src/services/awaySummary.ts

@@ -10,6 +10,9 @@ import { getSmallFastModel } from '../utils/model/model.js'
 import { asSystemPrompt } from '../utils/systemPromptType.js'
 import { getResolvedLanguage } from '../utils/language.js'
 import { queryModelWithoutStreaming } from './api/claude.js'
+import { createTrace, endTrace, isLangfuseEnabled } from './langfuse/index.js'
+import { getSessionId } from '../bootstrap/state.js'
+import { getAPIProvider } from '../utils/model/providers.js'
 import { getSessionMemoryContent } from './SessionMemory/sessionMemoryUtils.js'
 
 // Recap only needs recent context — truncate to avoid "prompt too long" on
@@ -42,6 +45,16 @@ export async function generateAwaySummary(
     return null
   }
 
+  const model = getSmallFastModel()
+  const langfuseTrace = isLangfuseEnabled()
+    ? createTrace({
+        sessionId: getSessionId(),
+        model,
+        provider: getAPIProvider(),
+        name: 'away-summary',
+      })
+    : null
+
   try {
     const memory = await getSessionMemoryContent()
     const recent = messages.slice(-RECENT_MESSAGE_WINDOW)
@@ -54,7 +67,7 @@ export async function generateAwaySummary(
       signal,
       options: {
         getToolPermissionContext: async () => getEmptyToolPermissionContext(),
-        model: getSmallFastModel(),
+        model,
         toolChoice: undefined,
         isNonInteractiveSession: false,
         hasAppendSystemPrompt: false,
@@ -62,6 +75,7 @@ export async function generateAwaySummary(
         querySource: 'away_summary',
         mcpTools: [],
         skipCacheWrite: true,
+        langfuseTrace,
       },
     })
 
@@ -69,14 +83,17 @@ export async function generateAwaySummary(
       logForDebugging(
         `[awaySummary] API error: ${getAssistantMessageText(response)}`,
       )
+      endTrace(langfuseTrace, undefined, 'error')
       return null
     }
+    endTrace(langfuseTrace)
     return getAssistantMessageText(response)
   } catch (err) {
     if (err instanceof APIUserAbortError || signal.aborted) {
       return null
     }
     logForDebugging(`[awaySummary] generation failed: ${err}`)
+    endTrace(langfuseTrace, undefined, 'error')
     return null
   }
 }

src/services/compact/compact.ts

@@ -1326,6 +1326,7 @@ async function streamCompactSummary({
           agents: context.options.agentDefinitions.activeAgents,
           mcpTools: [],
           effortValue: appState.effortValue,
+          langfuseTrace: context.langfuseTrace,
         },
       })
       const streamIter = streamingGen[Symbol.asyncIterator]()

src/services/tokenEstimation.ts

@@ -25,6 +25,8 @@ import { jsonStringify } from '../utils/slowOperations.js'
 import { isToolReferenceBlock } from '../utils/toolSearch.js'
 import { getAPIMetadata, getExtraBodyParams } from './api/claude.js'
 import { getAnthropicClient } from './api/client.js'
+import { createTrace, endTrace, isLangfuseEnabled, recordLLMObservation } from './langfuse/index.js'
+import { getSessionId } from '../bootstrap/state.js'
 import { withTokenCountVCR } from './vcr.js'
 
 // Minimal values for token counting with thinking enabled
@@ -309,6 +311,15 @@ export async function countTokensViaHaikuFallback(
       : betas
 
   // biome-ignore lint/plugin: token counting needs specialized parameters (thinking, betas) that sideQuery doesn't support
+  const apiStart = Date.now()
+  const langfuseTrace = isLangfuseEnabled()
+    ? createTrace({
+        sessionId: getSessionId(),
+        model: normalizeModelStringForAPI(model),
+        provider: getAPIProvider(),
+        name: 'token-estimation',
+      })
+    : null
   const response = await anthropic.beta.messages.create({
     model: normalizeModelStringForAPI(model),
     max_tokens: containsThinking ? TOKEN_COUNT_MAX_TOKENS : 1,
@@ -331,6 +342,22 @@ export async function countTokensViaHaikuFallback(
   const cacheCreationTokens = usage.cache_creation_input_tokens || 0
   const cacheReadTokens = usage.cache_read_input_tokens || 0
 
+  recordLLMObservation(langfuseTrace, {
+    model: normalizeModelStringForAPI(model),
+    provider: getAPIProvider(),
+    input: messagesToSend,
+    output: response.content,
+    usage: {
+      input_tokens: inputTokens,
+      output_tokens: usage.output_tokens,
+      cache_creation_input_tokens: cacheCreationTokens || undefined,
+      cache_read_input_tokens: cacheReadTokens || undefined,
+    },
+    startTime: new Date(apiStart),
+    endTime: new Date(),
+  })
+  endTrace(langfuseTrace)
+
   return inputTokens + cacheCreationTokens + cacheReadTokens
 }