mirror of
https://github.com/bleak-ai/gcontext.git
synced 2026-08-11 13:19:23 +02:00
bernatsamperaandClaude Opus 4.6
5fa3f6bffd
CLI command validates a local template folder against the workflow standard
(frontmatter, steps/, runs/example/), bundles files, and POSTs to the
marketplace API. --status mode queries GET /api/workflows/{id}/status.
16 new tests (suite 68 -> 84).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
workflows API
The marketplace backend for workflow templates. A FastAPI service with a Postgres database. Templates are stored as file bundles (one row per file) plus indexed manifest fields parsed server-side from the bundle's index.md frontmatter (see docs/workflows.md for the template standard).
Publishing is submit-for-review: anyone can POST a bundle, it lands pending, and it appears publicly only after approval.
Endpoints
GET /api/workflows: approved directory (id, name, description, tags).GET /api/workflows/{id}: one approved template with its full file bundle.POST /api/workflows: submit a bundle for review. Body:{"files": [{"path": "...", "content": "..."}]}.GET /api/moderation/workflows: pending entries (admin token).GET /api/moderation/workflows/{id}: one pending bundle (admin token).POST /api/moderation/workflows/{id}/approve: publish (admin token). Approving an id that is already published replaces the published content.POST /api/moderation/workflows/{id}/reject: hide (admin token).
Moderation auth is a single bearer token: Authorization: Bearer $ADMIN_TOKEN.
Configuration
Env vars: DATABASE_URL (postgresql+psycopg://...), ADMIN_TOKEN. Optional: MAX_FILE_BYTES (default 1000000), MAX_BUNDLE_BYTES (default 5000000), MAX_FILES (default 200).
Tables are created on startup.
Run locally
cd api
uv sync
DATABASE_URL=postgresql+psycopg://... ADMIN_TOKEN=dev uv run uvicorn app.main:app --reload
Tests
Tests need Docker (they start a throwaway Postgres container):
cd api
uv run pytest
Seed
ADMIN_TOKEN=... uv run python scripts/seed.py <template-folder> --api-url https://api.gcontext.ai