thedotmack/claude-mem · error · ServerClientError

invalid_response

invalid_response

Error message

sessionId is required for endSession

What it means

Thrown by ServerClient.endSession() when input.sessionId is falsy. The kind is 'invalid_response' (a client-side guard, despite the name) and fires before the POST /v1/sessions/<id>/end request is built, so no network call occurs on failure.

Source

Thrown at src/services/hooks/server-client.ts:222

    this.baseUrl = stripTrailingSlash(config.serverBaseUrl);
    this.apiKey = config.apiKey;
    this.timeoutMs = config.timeoutMs ?? DEFAULT_TIMEOUT_MS;
  }

  async startSession(input: ServerStartSessionRequest): Promise<ServerStartSessionResponse> {
    const body = this.buildStartSessionPayload(input);
    return this.request<ServerStartSessionResponse>('POST', '/v1/sessions/start', body);
  }

  async recordEvent(input: ServerRecordEventRequest): Promise<ServerRecordEventResponse> {
    const body = this.buildEventPayload(input);
    const path = input.generate === false ? '/v1/events?generate=false' : '/v1/events';
    return this.request<ServerRecordEventResponse>('POST', path, body);
  }

  async endSession(input: ServerEndSessionRequest): Promise<ServerEndSessionResponse> {
    if (!input.sessionId) {
      throw new ServerClientError('invalid_response', 'sessionId is required for endSession');
    }
    return this.request<ServerEndSessionResponse>(
      'POST',
      `/v1/sessions/${encodeURIComponent(input.sessionId)}/end`,
      {},
    );
  }

  // Phase 8 — direct observation insert (MCP `observation_add`). Calls
  // `/v1/memories`, which is the canonical write path that MUST NOT enqueue
  // a generation job. Anti-pattern guard for plan line 770: never duplicate
  // generation logic in MCP tools.
  async addObservation(
    input: ServerAddObservationRequest,
  ): Promise<ServerAddObservationResponse> {
    return this.request<ServerAddObservationResponse>(
      'POST',
      '/v1/memories',

View on GitHub (pinned to d768ba3643)

Solutions

  1. Capture and pass the sessionId returned by startSession() into endSession().
  2. Guard the end call: only invoke endSession when you have a non-empty sessionId.
  3. If session start failed, skip the end call entirely rather than passing an empty id.

Example fix

// before
await client.endSession({ sessionId: '' });
// throws ServerClientError(invalid_response, 'sessionId is required for endSession')

// after
const { sessionId } = await client.startSession({ ... });
// ... later, only if we got one
if (sessionId) {
  await client.endSession({ sessionId });
}
Defensive patterns

Strategy: validation

Validate before calling

if (!input?.sessionId) {
  throw new Error('endSession: sessionId is required (capture it from startSession)');
}

Type guard

function hasSessionId(input: unknown): input is { sessionId: string } {
  return typeof (input as any)?.sessionId === 'string' && (input as any).sessionId.length > 0;
}

Prevention

When it happens

Trigger: Calling endSession({ sessionId: '' }), endSession({}) (sessionId undefined), or with a null/whitespace id. The guard is a simple truthiness check on input.sessionId.

Common situations: Hook handler tries to end a session that was never started (no start response captured); sessionId variable was undefined after a failed startSession; caller passes the whole start response object instead of its id field.

Related errors


AI-assisted analysis of thedotmack/claude-mem@d768ba3643 (2026-08-12). Data as JSON: /api/errors/7a4dbfb606dfd749. Report an issue: GitHub.