feat: Fix observation timestamps, refactor session management, and enhance worker reliability (#437)

* Refactor worker version checks and increase timeout settings

- Updated the default hook timeout from 5000ms to 120000ms for improved stability.
- Modified the worker version check to log a warning instead of restarting the worker on version mismatch.
- Removed legacy PM2 cleanup and worker start logic, simplifying the ensureWorkerRunning function.
- Enhanced polling mechanism for worker readiness with increased retries and reduced interval.

* feat: implement worker queue polling to ensure processing completion before proceeding

* refactor: change worker command from start to restart in hooks configuration

* refactor: remove session management complexity

- Simplify createSDKSession to pure INSERT OR IGNORE
- Remove auto-create logic from storeObservation/storeSummary
- Delete 11 unused session management methods
- Derive prompt_number from user_prompts count
- Keep sdk_sessions table schema unchanged for compatibility

* refactor: simplify session management by removing unused methods and auto-creation logic

* Refactor session prompt number retrieval in SessionRoutes

- Updated the method of obtaining the prompt number from the session.
- Replaced `store.getPromptCounter(sessionDbId)` with `store.getPromptNumberFromUserPrompts(claudeSessionId)` for better clarity and accuracy.
- Adjusted the logic for incrementing the prompt number to derive it from the user prompts count instead of directly incrementing a counter.

* refactor: replace getPromptCounter with getPromptNumberFromUserPrompts in SessionManager

Phase 7 of session management simplification. Updates SessionManager to derive
prompt numbers from user_prompts table count instead of using the deprecated
prompt_counter column.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* refactor: simplify SessionCompletionHandler to use direct SQL query

Phase 8: Remove call to findActiveSDKSession() and replace with direct
database query in SessionCompletionHandler.completeByClaudeId().

This removes dependency on the deleted findActiveSDKSession() method
and simplifies the code by using a straightforward SELECT query.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* refactor: remove markSessionCompleted call from SDKAgent

- Delete call to markSessionCompleted() in SDKAgent.ts
- Session status is no longer tracked or updated
- Part of phase 9: simplifying session management

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* refactor: remove markSessionComplete method (Phase 10)

- Deleted markSessionComplete() method from DatabaseManager
- Removed markSessionComplete call from SessionCompletionHandler
- Session completion status no longer tracked in database
- Part of session management simplification effort

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* refactor: replace deleted updateSDKSessionId calls in import script (Phase 11)

- Replace updateSDKSessionId() calls with direct SQL UPDATE statements
- Method was deleted in Phase 3 as part of session management simplification
- Import script now uses direct database access consistently

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

* test: add validation for SQL updates in sdk_sessions table

* refactor: enhance worker-cli to support manual and automated runs

* Remove cleanup hook and associated session completion logic

- Deleted the cleanup-hook implementation from the hooks directory.
- Removed the session completion endpoint that was used by the cleanup hook.
- Updated the SessionCompletionHandler to eliminate the completeByClaudeId method and its dependencies.
- Adjusted the SessionRoutes to reflect the removal of the session completion route.

* fix: update worker-cli command to use bun for consistency

* feat: Implement timestamp fix for observations and enhance processing logic

- Added `earliestPendingTimestamp` to `ActiveSession` to track the original timestamp of the earliest pending message.
- Updated `SDKAgent` to capture and utilize the earliest pending timestamp during response processing.
- Modified `SessionManager` to track the earliest timestamp when yielding messages.
- Created scripts for fixing corrupted timestamps, validating fixes, and investigating timestamp issues.
- Verified that all corrupted observations have been repaired and logic for future processing is sound.
- Ensured orphan processing can be safely re-enabled after validation.

* feat: Enhance SessionStore to support custom database paths and add timestamp fields for observations and summaries

* Refactor pending queue processing and add management endpoints

- Disabled automatic recovery of orphaned queues on startup; users must now use the new /api/pending-queue/process endpoint.
- Updated processOrphanedQueues method to processPendingQueues with improved session handling and return detailed results.
- Added new API endpoints for managing pending queues: GET /api/pending-queue and POST /api/pending-queue/process.
- Introduced a new script (check-pending-queue.ts) for checking and processing pending observation queues interactively or automatically.
- Enhanced logging and error handling for better monitoring of session processing.

* updated agent sdk

* feat: Add manual recovery guide and queue management endpoints to documentation

---------

Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>

src/services/worker/SDKAgent.ts

@@ -115,6 +115,9 @@ export class SDKAgent {
           const discoveryTokens = (session.cumulativeInputTokens + session.cumulativeOutputTokens) - tokensBeforeResponse;
 
           // Process response (empty or not) and mark messages as processed
+          // Capture earliest timestamp BEFORE processing (will be cleared after)
+          const originalTimestamp = session.earliestPendingTimestamp;
+
           if (responseSize > 0) {
             const truncatedResponse = responseSize > 100
               ? textContent.substring(0, 100) + '...'
@@ -124,8 +127,8 @@ export class SDKAgent {
               promptNumber: session.lastPromptNumber
             }, truncatedResponse);
 
-            // Parse and process response with discovery token delta
-            await this.processSDKResponse(session, textContent, worker, discoveryTokens);
+            // Parse and process response with discovery token delta and original timestamp
+            await this.processSDKResponse(session, textContent, worker, discoveryTokens, originalTimestamp);
           } else {
             // Empty response - still need to mark pending messages as processed
             await this.markMessagesProcessed(session, worker);
@@ -145,8 +148,6 @@ export class SDKAgent {
         duration: `${(sessionDuration / 1000).toFixed(1)}s`
       });
 
-      this.dbManager.getSessionStore().markSessionCompleted(session.sessionDbId);
-
     } catch (error: any) {
       if (error.name === 'AbortError') {
         logger.warn('SDK', 'Agent aborted', { sessionId: session.sessionDbId });
@@ -254,19 +255,21 @@ export class SDKAgent {
   /**
    * Process SDK response text (parse XML, save to database, sync to Chroma)
    * @param discoveryTokens - Token cost for discovering this response (delta, not cumulative)
+   * @param originalTimestamp - Original epoch when message was queued (for backlog processing accuracy)
    */
-  private async processSDKResponse(session: ActiveSession, text: string, worker: any | undefined, discoveryTokens: number): Promise<void> {
+  private async processSDKResponse(session: ActiveSession, text: string, worker: any | undefined, discoveryTokens: number, originalTimestamp: number | null): Promise<void> {
     // Parse observations
     const observations = parseObservations(text, session.claudeSessionId);
 
-    // Store observations
+    // Store observations with original timestamp (if processing backlog) or current time
     for (const obs of observations) {
       const { id: obsId, createdAtEpoch } = this.dbManager.getSessionStore().storeObservation(
         session.claudeSessionId,
         session.project,
         obs,
         session.lastPromptNumber,
-        discoveryTokens
+        discoveryTokens,
+        originalTimestamp ?? undefined
       );
 
       // Log observation details
@@ -336,14 +339,15 @@ export class SDKAgent {
     // Parse summary
     const summary = parseSummary(text, session.sessionDbId);
 
-    // Store summary
+    // Store summary with original timestamp (if processing backlog) or current time
     if (summary) {
       const { id: summaryId, createdAtEpoch } = this.dbManager.getSessionStore().storeSummary(
         session.claudeSessionId,
         session.project,
         summary,
         session.lastPromptNumber,
-        discoveryTokens
+        discoveryTokens,
+        originalTimestamp ?? undefined
       );
 
       // Log summary details
@@ -422,6 +426,9 @@ export class SDKAgent {
       });
       session.pendingProcessingIds.clear();
 
+      // Clear timestamp for next batch (will be set fresh from next message)
+      session.earliestPendingTimestamp = null;
+
       // Clean up old processed messages (keep last 100 for UI display)
       const deletedCount = pendingMessageStore.cleanupProcessed(100);
       if (deletedCount > 0) {