# OpenRunner

> Governed ML platform for the full model lifecycle — datasets, experiment
> tracking + run instrumentation, training on cloud/serverless/self-hosted GPUs,
> hyperparameter sweeps, a container registry with serverless deployments, model
> cards, decisions (ADRs), handover, papers, and inference-drift monitoring. Ships
> a Python SDK + CLI (`openrunner`), an MCP server for AI coding tools, and
> `/openrunner:*` slash commands.

Read [llms-full.txt](https://openrun.gladia.io/llms-full.txt) for the complete,
self-contained end-to-end guide (every command + a new-project walkthrough). Quick
map below.

- Base URL: https://openrun.gladia.io · API: `/api/v1` · SDK: `pip install openrunner-sdk`
- Spine: **Org → Project → Run**. Datasets + registries are org-level; runs,
  deployments, decisions, papers are project-level. Refs are `org/name` or UUID.
- Governed: provider/registry secrets are stored encrypted server-side and never
  reach the client. Everything links (dataset → run → artifact → deployment).

## Lifecycle (see llms-full.txt for full commands)

- [Full guide](https://openrun.gladia.io/llms-full.txt): the entire doc inline.
- Auth & project: `openrunner login` · `openrunner init` · `/openrunner:new-project`.
- Datasets: `openrunner datasets create|upload|index|search|link` · `derive_by_query` · `/openrunner:new-dataset`, `/openrunner:derive-dataset`.
- Runs & instrumentation: `openrunner.init(project=, datasets=[...])` · `.log(...)` · `.log_artifact(...)` · model cards · `openrunner.log_decision(...)`.
- GPU training (governed, auto-terminate): `openrunner gpu run --provider runpod|vastai|lambda|ovhcloud|ovhai|modal|forge --project ORG/NAME "<cmd>"` · `openrunner gpu terminate` · `/openrunner:launch-gpu`, `/openrunner:train`, `/openrunner:forge`.
- Sweeps: `openrunner sweep create sweep.yaml` · `openrunner sweep agent SWEEP_ID` · `/openrunner:result-analyser`.
- Registry & deploy: `openrunner registry add|login` · `docker push` (auto-links through the broker) · `openrunner deploy link|images|create|scale|call` · `/openrunner:registry`, `/openrunner:deploy`, `/openrunner:build-api`.
- Ship: `openrunner handover check` · `/openrunner:handover` · `/openrunner:push-paper` · `/openrunner:drift-install`.

## AI-native

- MCP: `openrunner mcp` / `/openrunner:mcp` — read runs with `openrunner_watch_run`,
  `openrunner_get_run_metrics`, `openrunner_monitor_training`; knowledge via
  `openrunner_search_sessions|search_papers`; `openrunner_log_decision`;
  `openrunner_model_card_standard` / `openrunner_handover_standard`.
- SDK reference: mirrors `openrunner/CLAUDE.md` in the repo.

## Optional

- [New project, end to end](https://openrun.gladia.io/llms-full.txt): §9 of the full
  guide — create → data → train → sweep → package → registry → deploy → handover.
