---

name: gerev description: "Route Gerev tasks across connector setup, search/indexing, and source or Docker runtime workflows." disable-model-invocation: true metadata: disco-role: operating license: MIT

Gerev

Gerev is a FastAPI-backed enterprise search service with pluggable data sources, a document indexing pipeline, and a React UI. Use this skill as the router for future Gerev tasks; do not stop at root guidance when a deeper sub-skill owns the workflow.

Route map

• data-source-connectors — connector discovery, validation, location selection, add/remove flows, and data-source setup docs.
• search-indexing — query ranking, Faiss/BM25 indexing, parser helpers, model loading, queues, search/status routes, and startup-time search failures.
• deployment-runtime — source startup, Docker/compose, UI build, storage layout, ports, and backend/UI boot troubleshooting.

Install and first checks

• Backend runtime lives under app/. Install runtime dependencies with:

  cd app
  python -m pip install -r requirements.txt
  

• If the resolver pulls newer major versions that break imports, keep the compatibility family that matches the current code: fastapi==0.95.2, starlette==0.27.0, pydantic<2, transformers<5, and sentence-transformers<4.

• Frontend build lives under ui/. Use cd ui && npm install && npm run build when the UI bundle is needed for Docker or source serving.

• For a read-only connector inventory without importing the heavy app stack, run:

  python sub-skills/data-source-connectors/scripts/inspect_data_sources.py --app-dir app --json
  

  The bundled connector helper is inspect_data_sources.py.

Current repo caveat

A full backend import or startup may fail with ImportError: cannot import name 'split_PDF_into_paragraphs' from app/parsers/pdf.py because app/indexing/index_documents.py imports a symbol that no longer exists. Treat that as a troubleshooting item, not an environment success. See references/troubleshooting.md and the search-indexing sub-skill before claiming a clean boot.

When to read shared references

• Read repo-provenance.md when checking whether this skill still matches the current checkout or before a refresh.
• Read troubleshooting.md for cross-cutting install/import, model-cache, tokenizer, storage, and startup failures.
• Read frontend-api.md when you need the backend route contracts that the UI calls.
• Read deployment-runtime.md when you need the source-start, Docker, compose, or storage layout details.
• Read the nearest sub-skills/<id>/references/ files for workflow-specific depth that the root router intentionally leaves out.

Minimal read-only checks

• Connector discovery only: python sub-skills/data-source-connectors/scripts/inspect_data_sources.py --app-dir app --strict
• Search stack or model-cache check: use sub-skills/search-indexing/scripts/inspect_search_indexing.py when you need ranking/runtime details.
• Data-source catalog check: use inspect_data_sources.py when you need connector schema or UI flows.
• Source-start/runtime layout check: use inspect_runtime_paths.py when you need Docker/source boot details.

Use this root skill when...

• The user says Gerev, enterprise search, add a connector, configure a data source, inspect search results, debug indexing, start the backend, or build the UI.
• The task spans more than one sub-skill and you need the router to point to the correct deeper reference instead of guessing from source folders.