Config
AI review options
Optional plan annotations only. AI is not part of the deterministic rewrite path.
| Option | Default | Description |
|---|---|---|
--ai-review | false | Prepare AI review for the generated plan. |
--ai-dry-run | false | Write the request without calling a provider. |
--ai-provider <provider> | mock | Provider: mock, openai, openai-compatible, anthropic, or ollama. mock reads a local response file and does not call a network API. |
--ai-context <mode> | snippets | Context mode: summary, snippets, or full. |
--ai-request-out <path> | none | Path for AI request JSON. |
--ai-response-file <path> | required for mock | Local JSON decisions for the mock provider. Required when mock actually runs. |
--ai-endpoint <url> | provider default | Required for openai-compatible. OpenAI, Anthropic, and Ollama use their own defaults when omitted. |
--ai-model <model> | required except mock | Required for openai, openai-compatible, anthropic, and ollama. |
--ai-api-key-env <name> | provider default | Environment variable containing the API key. |
--ai-timeout-ms <number> | 60000 | Request timeout in milliseconds. |
--ai-on-error <mode> | fail | Failure mode: fail, skip, or mark-needs-review. |
--ai-cache <path> | none | AI review cache JSON path. |
--ai-max-candidates <number> | none | Maximum candidates sent for review. |
--ai-max-context-bytes <number> | none | Maximum source context bytes sent for review. |
Inspect the request before sending
Use a dry run to write the provider request without making a network call. Review the JSON and remove source that contains secrets or other data the provider must not receive.
shell
pnpm exec type-tune plan \
--project tsconfig.json \
--ai-review \
--ai-dry-run \
--ai-request-out .type-tune/ai-request.json