Worker crashes on startup due to unhandled Promise rejection in ensureBackfilled() (v8.0.2 regression)

[hayaishi]

Worker crashes on startup due to unhandled Promise rejection in ensureBackfilled() (v8.0.2 regression)

Bug Description

The worker service crashes during startup in v8.0.2 due to an unhandled Promise rejection in ChromaSync.ensureBackfilled(). This is a regression from v7.4.5 where the error was properly caught and logged as non-fatal.

Root Cause: The .catch() error handler was removed from DatabaseManager.ts in v8.0.2, causing any errors during Chroma backfill to crash the worker process instead of being logged as warnings.

Steps to Reproduce

1. Install claude-mem v8.0.2 via claude plugin add claude-mem
2. Start the worker service (happens automatically or via npm run worker:start)
3. Worker crashes during Chroma backfill process when encountering any error

Common triggers:

• Chroma MCP server returns malformed JSON responses
• Network issues during Chroma communication
• Chroma collection initialization errors
• Python/uvx environment issues

Expected Behavior

Worker should start successfully and log any Chroma sync errors as non-fatal warnings (as in v7.4.5). The worker can operate without Chroma, falling back to SQLite FTS search.

Environment

• Claude-mem version: 8.0.2
• Claude Code version: (not applicable - installed via plugin)
• OS: macOS (Darwin 25.1.0)
• Platform: darwin
• Node.js: v24.11.1
• Bun: 1.3.5
• Python: 3.11.9 (system), 3.13 (configured in settings)
• uv: 0.9.15

Logs

Worker logs are located at:

• Path: ~/.claude-mem/logs/worker-YYYY-MM-DD.log
• Example: ~/.claude-mem/logs/worker-2025-12-23.log

Please paste relevant log entries (last 50 lines or error messages):

[2025-12-23 13:41:00.180] [INFO ] [CHROMA_SYNC] Connected to Chroma MCP server {project=claude-mem}
[2025-12-23 13:41:00.182] [INFO ] [CHROMA_SYNC] Fetching existing Chroma document IDs... {project=claude-mem}
[2025-12-23 13:41:00.183] [ERROR] [CHROMA_SYNC] Failed to fetch existing IDs {project=claude-mem} JSON Parse error: Unexpected identifier "Error"
717 |         created_at,
718 |         created_at_epoch
719 |       FROM user_prompts
720 |       WHERE claude_session_id = ?
721 |       ORDER BY prompt_number ASC
722 |     `).all(e)}close(){this.db.close()}};Iu();mt();Dr();dr();var r1=kt(require("path"),1),a1=kt(require("os"),1),Nu=class{client=null;transport=null;connected=!1;project;collectionName;VECTOR_DB_DIR;BATCH_SIZE=100;constructor(e){this.project=e,this.collectionName=`cm__${e}`,this.VECTOR_DB_DIR=r1.default.join(a1.default.homedir(),".claude-mem","vector-db")}async ensureConnection(){if(!(this.connected&&this.client)){U.info("CHROMA_SYNC","Connecting to Chroma MCP server...",{project:this.project});try{let r=ct.loadFromFile(In).CLAUDE_MEM_PYTHON_VERSION,n=process.platform==="win32",a={command:"uvx",args:["--python",r,"chroma-mcp","--client-type","persistent","--data-dir",this.VECTOR_DB_DIR],stderr:"ignore"};n&&(a.windowsHide=!0,U.debug("CHROMA_SYNC","Windows detected, attempting to hide console window",{project:this.project})),this.transport=new Ms(a),this.client=new As({name:"claude-mem-chroma-sync",version:"1.0.0"},{capabilities:{}}),await this.client.connect(this.transport),this.connected=!0,U.info("CHROMA_SYNC

SyntaxError: JSON Parse error: Unexpected identifier "Error"
      at getExistingChromaIds (/Users/akihiro.hayaishi/.claude/plugins/marketplaces/thedotmack/plugin/scripts/worker-service.cjs:722:6756)
      at async ensureBackfilled (/Users/akihiro.hayaishi/.claude/plugins/marketplaces/thedotmack/plugin/scripts/worker-service.cjs:722:7495)

Bun v1.3.5 (macOS arm64)

Additional Context

Code Regression Analysis

Location: src/services/worker/DatabaseManager.ts (lines 30-33)

v7.4.5 (working):

// Start background backfill (fire-and-forget, with error logging)
this.chromaSync.ensureBackfilled().catch((error) => {
  logger.error('DB', 'Chroma backfill failed (non-fatal)', {}, error);
});

v8.0.2 (broken):

// Start background backfill (fire-and-forget)
this.chromaSync.ensureBackfilled();

Impact

When ensureBackfilled() encounters any error, the Promise rejection goes unhandled, causing the worker process to crash. This prevents the worker from starting successfully and makes the plugin completely unusable.

Proposed Fix

Restore the .catch() error handler in DatabaseManager.ts:

async initialize(): Promise<void> {
  this.sessionStore = new SessionStore();
  this.sessionSearch = new SessionSearch();

  // Initialize ChromaSync
  this.chromaSync = new ChromaSync('claude-mem');

  // Start background backfill (fire-and-forget, with error logging)
  this.chromaSync.ensureBackfilled().catch((error) => {
    logger.error('DB', 'Chroma backfill failed (non-fatal)', {}, error);
  });

  logger.info('DB', 'Database initialized');
}

This ensures that Chroma sync errors are logged but don't prevent the worker from starting, maintaining backward compatibility with v7.4.5 behavior.

Workaround

Currently, the only workaround is to:

1. Downgrade to v7.4.5, or
2. Build from the v7.4.5 tag manually

Reference

• Git diff: v7.4.5...v8.0.2
  • 3ea0b60#diff-5a10dd80f449c1a669c97435c404e26638eb127a7b96bea860f297c6a65bb45fL33-R34
• The error handler removal appears to have been unintentional
• In v7.4.5, the comment explicitly states "with error logging" and treats Chroma backfill failures as non-fatal, which is correct since the worker can operate without Chroma

[github-actions]

In version 8.0.2 of the claude-mem plugin, the worker crashes on startup due to an unhandled Promise rejection in the ChromaSync.ensureBackfilled() function, which previously had an error handler in version 7.4.5 that logged errors without causing a crash. This regression stems from the accidental removal of the .catch() handler in DatabaseManager.ts, resulting in critical failures whenever Chroma synchronization encounters issues such as malformed JSON responses, network problems, or environment misconfigurations. The issue prevents the worker from fully starting, making the plugin unusable, whereas earlier versions allowed the worker to fall back to SQLite search in such cases. A proposed fix involves reintroducing the .catch() error handler to log errors as non-fatal while maintaining operational stability. Until this is resolved, users are advised to downgrade to version 7.4.5 or manually apply the fix.