Service API knowledge pipeline endpoints under /v1 return 404

[tokezooo]

Summary

Service API endpoints for Knowledge Pipeline (RAG pipeline) are not reachable under /v1 and return 404, even though the controller exists in the codebase.

Steps to reproduce

1. Self-host Dify and create a dataset with Knowledge Pipeline enabled
2. Call the Service API endpoint:

curl -i -H "Authorization: Bearer $DIFY_API_KEY" \
  "$DIFY_BASE_URL/v1/datasets/<dataset_id>/pipeline/datasource-plugins"

Expected behavior

• Endpoint exists and returns 200 with a JSON array (or a non-404 error if request is invalid).

Actual behavior

• Returns 404 Not Found.

Notes / suspected root cause

• api/controllers/service_api/dataset/rag_pipeline/rag_pipeline_workflow.py exists but is not imported/registered in the Service API blueprint aggregation module (api/controllers/service_api/__init__.py), so routes never get registered.
• Route decorators in that module currently use invalid patterns like f"/datasets/{uuid:dataset_id}/..." instead of Flask converter syntax "/datasets/<uuid:dataset_id>/...". If the module is imported, this can also break import/init.

[dosubot]

Hi @tokezooo! I'm Dosu and I’m helping the dify team.

The 404 error happens because the Knowledge Pipeline (RAG pipeline) Service API endpoints are not registered in the Service API blueprint. The controller file exists, but it is not imported in api/controllers/service_api/__init__.py, so its routes are never exposed. Even if you import it, the route decorators use invalid syntax like f"/datasets/{uuid:dataset_id}/..." instead of Flask's "/datasets/<uuid:dataset_id>/...", which prevents route registration and can cause import errors. Both issues must be fixed for the endpoints to work as expected [source] [source].

To fix this:

1. Update all route decorators in rag_pipeline_workflow.py to use valid Flask syntax, for example:
   Change

   @service_api_ns.route(f"/datasets/{uuid:dataset_id}/pipeline/datasource-plugins")

   to

   @service_api_ns.route("/datasets/<uuid:dataset_id>/pipeline/datasource-plugins")

   Do this for every route in the file.

2. Import the controller in api/controllers/service_api/__init__.py so the routes are registered:

   from .dataset.rag_pipeline import rag_pipeline_workflow

Be aware that these endpoints are not officially released or fully supported yet, so you may encounter other issues or limitations. The most reliable way to trigger Knowledge Pipeline workflows is still through the Web UI for now [source] [source].

If you want to track progress or future support for these endpoints, keep an eye on related issues and discussions in the Dify repository.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.