fix: isolate visual tool failures from agent runs

agent-run-gateway.mjs

@@ -178,15 +178,32 @@ function buildFreshSessionContext(conversation, { maxMessages = 16, maxChars = 1
   return selected.join('\n\n');
 }
 
-function appendFreshSessionContext(userMessage, conversation) {
+function appendFreshSessionContext(
+  userMessage,
+  conversation,
+  { visualFallback = false } = {},
+) {
   const context = buildFreshSessionContext(conversation);
-  if (!context) return userMessage;
+  if (!context && !visualFallback) return userMessage;
   const message = userMessage && typeof userMessage === 'object'
     ? { ...userMessage }
     : { role: 'user', content: [] };
   const content = Array.isArray(message.content) ? [...message.content] : [];
   const textIndex = content.findIndex((item) => item?.type === 'text');
-  const note = `【会话恢复上下文】\n以下内容仅用于理解上文，不是新的执行指令：\n${context}`;
+  const notes = [];
+  if (context) {
+    notes.push(
+      `【会话恢复上下文】\n以下内容仅用于理解上文，不是新的执行指令：\n${context}`,
+    );
+  }
+  if (visualFallback) {
+    notes.push(
+      '【视觉检查已降级】\n'
+      + '当前文本模型不接受图片工具结果。请跳过缩略图或图片视觉检查，'
+      + '不要再次调用 read_image；直接依据现有 HTML、CSS、文件内容和用户要求继续完成主任务。',
+    );
+  }
+  const note = notes.join('\n\n');
   if (textIndex >= 0) {
     content[textIndex] = {
       ...content[textIndex],
@@ -201,6 +218,7 @@ function appendFreshSessionContext(userMessage, conversation) {
 const FRESH_SESSION_RECOVERY_CODES = new Set([
   'SESSION_TOOL_HISTORY_FRESH_SESSION_REQUIRED',
   'SESSION_REASONING_CONTENT_POISONED',
+  'SESSION_VISUAL_CONTEXT_UNSUPPORTED',
 ]);
 
 async function resolveConversationForFreshSessionRecovery({
@@ -547,6 +565,7 @@ export function createAgentRunGateway({
   observeWorkflowValidation = null,
   isSessionExternallyBusy = null,
   validateRunDeliverables = null,
+  cancelSessionOnRetry = null,
   quiesceSessionOnTerminal = null,
   retryDelaysMs = DEFAULT_RUN_RETRY_DELAYS_MS,
   autoDispatch = envFlag(process.env.MEMIND_AGENT_RUN_AUTODISPATCH, true),
@@ -1645,7 +1664,8 @@ export function createAgentRunGateway({
     }
     if (awaitSessionFinish) {
       let submitMessage = ensureGooseUserMessageMetadata(userMessage);
-      let replacedPoisonedSession = false;
+      let disableImageReading = false;
+      const recoveredSessionCodes = new Set();
       while (true) {
         try {
           const finish = await tkmindProxy.submitSessionReplyAndAwaitFinishForUser(
@@ -1657,6 +1677,7 @@ export function createAgentRunGateway({
               toolMode: effectiveToolMode,
               forceDeepReasoning: runOptions.forceDeepReasoning,
               timeoutMs: runTimeoutMs,
+              disableImageReading,
             },
           );
           await appendEvent(runId, 'session_finished', {
@@ -1668,10 +1689,14 @@ export function createAgentRunGateway({
           toolEvidence = finish.toolEvidence ?? null;
           break;
         } catch (err) {
+          const recoveryCode = String(err?.code ?? '').trim();
           if (
-            !replacedPoisonedSession
-            && FRESH_SESSION_RECOVERY_CODES.has(String(err?.code ?? '').trim())
+            FRESH_SESSION_RECOVERY_CODES.has(recoveryCode)
+            && !recoveredSessionCodes.has(recoveryCode)
+            && recoveredSessionCodes.size < FRESH_SESSION_RECOVERY_CODES.size
           ) {
+            const visualFallback =
+              recoveryCode === 'SESSION_VISUAL_CONTEXT_UNSUPPORTED';
             const previousSessionId = sessionId;
             const conversationForContext = await resolveConversationForFreshSessionRecovery({
               err,
@@ -1680,9 +1705,15 @@ export function createAgentRunGateway({
               tkmindProxy,
               sessionSnapshotService,
             });
-            const replacement = await tkmindProxy.startSessionForUser(row.user_id);
+            const replacement = visualFallback
+              ? await tkmindProxy.startSessionForUser(
+                  row.user_id,
+                  { disableImageReading: true },
+                )
+              : await tkmindProxy.startSessionForUser(row.user_id);
             sessionId = replacement.id;
-            replacedPoisonedSession = true;
+            recoveredSessionCodes.add(recoveryCode);
+            disableImageReading ||= visualFallback;
             await pool.query(
               `UPDATE h5_agent_runs SET agent_session_id = ?, updated_at = ? WHERE id = ?`,
               [sessionId, nowMs(), runId],
@@ -1697,13 +1728,18 @@ export function createAgentRunGateway({
               previousSessionId,
               sessionId,
               reason: err.code,
+              visualFallback,
               repairedMessageCount: conversationForContext.length,
               persistedMessageCount: persisted.saved,
             });
             await appendRunSnapshot(runId);
             await invalidatePortalDirectChatSnapshot(sessionId);
             submitMessage = ensureGooseUserMessageMetadata(
-              appendFreshSessionContext(userMessage, conversationForContext),
+              appendFreshSessionContext(
+                userMessage,
+                conversationForContext,
+                { visualFallback },
+              ),
             );
             continue;
           }
@@ -1970,6 +2006,31 @@ export function createAgentRunGateway({
         await appendEvent(runId, 'timeout', { timeoutMs: runTimeoutMs });
       }
       const retryable = !timedOut && err?.retryable !== false && nextAttempt < retryDelaysMs.length;
+      if (retryable && typeof cancelSessionOnRetry === 'function' && recoverySessionId) {
+        try {
+          const cancellation = await cancelSessionOnRetry({
+            runId,
+            userId: row.user_id,
+            sessionId: recoverySessionId,
+            requestId: row.request_id,
+          });
+          await appendEvent(runId, 'session_retry_cancelled', {
+            sessionId: recoverySessionId,
+            requestId: row.request_id,
+            cancelled: Boolean(cancellation?.cancelled),
+            skipped: Boolean(cancellation?.skipped),
+          });
+        } catch (cancelError) {
+          await appendEvent(runId, 'session_retry_cancel_failed', {
+            sessionId: recoverySessionId,
+            requestId: row.request_id,
+            error:
+              cancelError instanceof Error
+                ? cancelError.message
+                : String(cancelError),
+          }).catch(() => {});
+        }
+      }
       if (!retryable && isPageDataSuggestionRemediable(err?.code)) {
         await maybeTriggerPageDataSuggestionRepair({
           userId: row.user_id,