[waleedlatif1]

Summary

• Hosted fal generations are billed at completion from fal's per-request receipt when it is ready, else from fal's per-call historical estimate or a fixed floor when the estimate is rate-limited. fal records the receipt minutes after the request completes, so most generations were billed on the estimate, and token-priced models (Seedance) often cost several times the per-call average
• New falai-cost-settlement Trigger task: a generation billed on an estimate schedules a settlement (2 min delay, retried with backoff for about an hour) that reads fal's billing event for the request and records the shortfall as one ledger row. The row's event key is the fal request id, so retries never bill twice; an estimate above the exact charge stays billed, since the ledger records charges only
• Wired for the workflow/API hosted tools (video_falai, image_generate, through a declarative hosting.exactCostSettlement: 'falai' read by executeTool, at the tool's markup) and Chat's generate_video/generate_audio (at cost, matching how Mothership bills fal media). The hosted key travels as its env var name, never its value, so the task reads the key that created the request
• The video and image fal poll loops now cancel the queued request when they stop waiting (timeout, abort, status error) before fal reports a terminal status, so an abandoned request is not run and charged; this reuses the existing best-effort cancelFalRequest
• Settlement runs only on hosted, billing-enabled deployments with Trigger.dev configured

Type of Change

• Bug fix

Testing

• lib/tools/falai-settlement.integration.ts (real Postgres, fal billing-events stubbed): the shortfall is billed exactly once across repeated settlements (fails with a per-attempt ledger key), nothing is billed while fal has no receipt, and a missing hosted key refuses without calling fal
• lib/internal/image/operations.test.ts: polling that stops before completion cancels the fal request (fails without the cancel)
• lib/media/falai.test.ts, lib/mothership/tools/server/media/*, lib/internal/video, tools/index.test.ts: existing suites updated for the cost recorder's hosted-key argument
• biome, check:unused-exports, check:test-patterns, check:source-conventions, check:comment-hygiene, check:utils, check:explicit-any, check:file-names, check:boundaries, check:client-boundary, check:tool-registry-boundary; type-check and full audits in CI

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Nov 6, 2026 8:17pm UTC

[greptile-apps]

Confidence Score: 5/5

[Critical impact] The reviewed changes appear safe to merge.

Summary

Adds delayed settlement for hosted fal generations initially billed from estimates.

• Hosted Fal.ai generations settle to their exact cost.
• Fal.ai image and video requests are cancelled when polling stops early.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Hosted fal generation completes] --> B{Exact receipt ready?}
  B -->|Yes| C[Bill receipt cost]
  B -->|No| D[Bill estimate, including zero]
  D --> E[Schedule settlement after two minutes]
  E --> F{Receipt ready?}
  F -->|No| G[Retry with backoff]
  G --> F
  F -->|Yes| H{Exact cost exceeds estimate?}
  H -->|No| I[Leave original charge unchanged]
  H -->|Yes| J[Refresh billing period if it ended]
  J --> K[Record shortfall with stable event key]

Loading

_{Reviews (4) · Last reviewed commit: "fix(billing): settle zero fal estimates ..." · Reviewed by Greptile}

[cubic-dev-ai]

All reported issues were addressed across 19 files

_{Turn on auto-fix | Re-trigger cubic}

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

All reported issues were addressed across 19 files

<sub>Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 20 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Turn on auto-fix | Re-trigger cubic}