# Sutro Docs

## Docs

- [Sutro](https://docs.sutro.sh/index.md): Build AI Functions from expert feedback, then run them at scale.
- [Build your first Function](https://docs.sutro.sh/quickstart.md): Teach Sutro a binary decision from a representative dataset.
- [Design a Function](https://docs.sutro.sh/functions/build.md): Choose a narrow decision, output shape, and representative dataset.
- [Improve a Function](https://docs.sutro.sh/functions/improve.md): Use disagreement, annotations, and held-out examples to refine behavior.
- [Run a Function](https://docs.sutro.sh/functions/run.md): Invoke a published Function through Sutro Batch or copy its prompt.
- [Run Functions at scale](https://docs.sutro.sh/batch/run-model.md): Execute published Functions over production datasets with Sutro Batch.
- [Production Batch](https://docs.sutro.sh/batch/production.md): Estimate, submit, monitor, and download large Batch jobs.
- [Current capabilities](https://docs.sutro.sh/reference/current-capabilities.md): What Sutro supports today and where important limits remain.
- [Deployment](https://docs.sutro.sh/reference/deployment.md): Choose how Sutro Functions, Batch, and real-time serving run.
- [Authentication](https://docs.sutro.sh/reference/authentication.md): Configure a Sutro deployment URL and API key.
- [MCP](https://docs.sutro.sh/reference/mcp.md): Let an agent drive the Sutro Function workflow with your feedback.
- [Python SDK](https://docs.sutro.sh/reference/python-sdk/setup.md): Install the Sutro SDK and choose an interface.
- [Functions](https://docs.sutro.sh/reference/python-sdk/functions.md): Run a published text Function through Sutro Batch.
- [Direct model inference](https://docs.sutro.sh/reference/python-sdk/batch-inference.md): Call one or more model IDs directly instead of using a published Function.
- [Remote inputs](https://docs.sutro.sh/reference/python-sdk/presigned-s3-inputs.md): Submit production CSV or Parquet inputs by signed HTTPS URL.
- [Jobs and results](https://docs.sutro.sh/reference/python-sdk/job-methods.md): Monitor jobs and retrieve, download, list, or cancel results.
- [CLI](https://docs.sutro.sh/reference/cli/index.md): Authenticate, inspect jobs, retrieve results, and manage the local cache.
- [Run a Function in Batch](https://docs.sutro.sh/reference/api/running-batch-inference.md): Execute a published Function asynchronously over inline or remote inputs.
- [Check job status](https://docs.sutro.sh/reference/api/job-status.md): Read the current state of a Batch job.
- [Fetch a job](https://docs.sutro.sh/reference/api/fetch-job.md): Retrieve one Batch job and its metadata.
- [Cancel a job](https://docs.sutro.sh/reference/api/cancel-job.md): Request cancellation of an active Batch job.
- [Download job results](https://docs.sutro.sh/reference/api/job-results.md): Download a completed Batch job as JSON, CSV, or Parquet.
- [Get a results URL](https://docs.sutro.sh/reference/api/job-results-url.md): Create or reuse a unified Parquet artifact and return presigned download URLs.
- [List jobs](https://docs.sutro.sh/reference/api/list-jobs.md): List Batch jobs, newest first.

## OpenAPI Specs

- [openapi](https://docs.sutro.sh/api-reference/openapi.json)