> ## Documentation Index
> Fetch the complete documentation index at: https://docs.r3al.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Project workspace

> Session-authenticated endpoints for team-owned setup, data, and models.

Workspace endpoints require a signed-in session. Admins and developers can create and edit; members can read. Team identity is taken from the session, never from a request body. Project inference has a separate [API-key entry point](/api/inference).

## Projects

| Method | Endpoint | Behavior |
| - | - | - |
| GET | `/v1/project-starters` | Read the starter catalog and provenance. |
| GET | `/v1/projects` | List team projects and resumable conversations. |
| POST | `/v1/projects` | Create a draft specification. |
| GET | `/v1/projects/{id}` | Read specification, registered models, and runtime state. |
| PUT | `/v1/projects/{id}` | Update with `{ "version": n, "spec": { ... } }`. |
| POST | `/v1/projects/{id}/plan` | Review a version with `{ "version": n }`. |

## Datasets

Under `/v1/projects/{id}/datasets`: `GET` lists datasets and `POST` imports a starter or records. A specific `/{dataset_id}` supports `GET` and versioned `PUT`. `POST /{dataset_id}/freeze` freezes a validated version; `POST /{dataset_id}/fork` makes an editable copy.

Imports are bounded to 5 MB, 500 records, and 50 fields. Records contain `id`, `group`, `input`, and `expected`. Company record import requires managed preparation/training. See the [data guide](/platform/datasets).

## Models

Under `/v1/projects/{id}/models`: `GET` lists models and `POST` creates a draft with a UUID `request_id` and `spec`. A specific `/{model_id}` supports `GET` and versioned `PUT`. `POST /{model_id}/register` registers the expected version.

See [model registration](/platform/models) for the configuration fields and connection requirements.

## Version conflicts

Edits and review operations use the last-read `version`. Stale writes return HTTP 409. Read the latest object before applying another change. Frozen datasets and registered models are immutable; create a new version instead of overwriting them.

These APIs save project objects. They do not dispatch training, pruning, quantization, or fine-tuning jobs.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.