What causes "failed to create flow tool call log: %w" in vxcontrol/pentagi?

Inspect the wrapped error and Postgres logs for the failing statement

failed to create flow tool call log: %w

Error message

failed to create flow tool call log: %w

What it means

Sixth step of newFlowProviderWorkers: the tool call log worker is created via cnts.tclc.NewFlowToolCallLog(ctx, flowID, pub). An error wrapped with this message aborts flow startup, because agent tool-call events could not be recorded persistently.

Source

Thrown at backend/pkg/controller/flow.go:1194

	slw, err := cnts.slc.NewFlowSearchLog(ctx, flowID, pub)
	if err != nil {
		return nil, fmt.Errorf("failed to create flow search log: %w", err)
	}

	tlw, err := cnts.tlc.NewFlowTermLog(ctx, flowID, pub)
	if err != nil {
		return nil, fmt.Errorf("failed to create flow term log: %w", err)
	}

	vslw, err := cnts.vslc.NewFlowVectorStoreLog(ctx, flowID, pub)
	if err != nil {
		return nil, fmt.Errorf("failed to create flow vector store log: %w", err)
	}

	tclw, err := cnts.tclc.NewFlowToolCallLog(ctx, flowID, pub)
	if err != nil {
		return nil, fmt.Errorf("failed to create flow tool call log: %w", err)
	}

	sw, err := cnts.sc.NewFlowScreenshot(ctx, flowID, pub)
	if err != nil {
		return nil, fmt.Errorf("failed to create flow screenshot: %w", err)
	}

	return &flowProviderWorkers{
		mlw:  mlw,
		alw:  alw,
		slw:  slw,
		tlw:  tlw,
		vslw: vslw,
		tclw: tclw,
		sw:   sw,
	}, nil
}

View on GitHub (pinned to ea665308ba)

Solutions

Inspect the wrapped error and Postgres logs for the failing statement
Run pending migrations so the tool-call-log schema exists
Increase DB pool size or start flows with less concurrency
Retry flow creation once DB health is confirmed

Example fix

// before: one slow step eats the whole startup budget
// (ctx shared, no deadline management)
tclw, err := cnts.tclc.NewFlowToolCallLog(ctx, flowID, pub)
// after: give worker creation its own deadline
wctx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 30*time.Second)
defer cancel()
tclw, err = cnts.tclc.NewFlowToolCallLog(wctx, flowID, pub)
if err != nil {
    return nil, fmt.Errorf("failed to create flow tool call log (flow=%d): %w", flowID, err)
}

Defensive patterns

Strategy: retry

Validate before calling

var exists int
if err := db.QueryRowContext(ctx, `SELECT COUNT(1) FROM flows WHERE id=$1`, flowID).Scan(&exists); err != nil || exists == 0 {
    return fmt.Errorf("flow %d gone; cannot create tool call log", flowID)
}

Type guard

func hasValidFlowID(flowID int64) bool { return flowID > 0 }

Try / catch

tclw, err := cnts.tclc.NewFlowToolCallLog(ctx, flowID, pub)
if err != nil {
    if isTransientDBError(err) { /* retry with exponential backoff */ }
    return nil, fmt.Errorf("failed to create flow tool call log (flow=%d): %w", flowID, err)
}

Prevention

Raise the DB connection pool limit if many flows start simultaneously
Run migrations on every deploy
Inspect the wrapped cause before assuming the flow is unrecoverable
Use context.WithTimeout per step so earlier slow steps do not exhaust the budget

When it happens

Trigger: NewFlowToolCallLog fails during flow startup — DB error creating the tool-call log for this flow, invalid/deleted flowID, or context cancelled during the sequential worker-creation chain.

Common situations: PostgreSQL connection pool exhausted during many simultaneous flow starts; tool_call_logs table missing due to skipped migration; ctx deadline already exceeded after earlier slow steps; concurrent flow deletion.

Related errors

AI-assisted analysis of vxcontrol/pentagi@ea665308ba (2026-09-01). Data as JSON: /api/errors/e8bc159eeb8dd878. Report an issue: GitHub.