Scope decision: file search / provider-hosted RAG as a built-in tool?
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 176
- Forks
- 28
- Avg merge
- 18h 42m
- Merged PRs (30d)
- 16
Description
OpenAI and Google both ship a GA, fully provider-hosted RAG tool. Should chatlas expose it as a first-class built-in (tool_file_search()), or is it out of scope?
| Provider | Tool spec | Notes |
|---|---|---|
| OpenAI (Responses) | {"type": "file_search", "vector_store_ids": [...]} |
$2.50/1k calls + $0.10/GB/day storage (first GB free). Docs |
types.Tool(file_search=types.FileSearch(file_search_store_names=[...])) |
Gemini API only — not supported on Vertex AI. Embeddings billed at indexing; storage free. Docs | |
| Anthropic | — | No equivalent (closest is the Files API + code execution). |
The case for
- Pairs naturally with the provider Files API work (#334) — "upload docs, then let the model search them" is a very common ask.
- Entirely provider-hosted: chatlas would only pass store IDs, not implement retrieval.
The case against
- #360 explicitly lists "no RAG primitives, vector stores, or document loaders" as a non-goal. Even if chatlas only passes IDs, this feature creates gravitational pull toward vector-store lifecycle helpers (create store, upload, poll indexing status, metadata filters) — that's where the real UX friction lives, and it's exactly the surface #360 declines.
- Only two of three first-class providers, and the Google variant doesn't work on Vertex.
- The
ToolBuiltInescape hatch already works today for users who manage their own vector stores.
Like the other (decision) issues from #360, closing this as "no, use ToolBuiltIn + a docs recipe" is a legitimate resolution.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Read the scope discussion in #360 and the provider Files API work in #334, then inspect the existing ToolBuiltIn escape hatch. Compare the linked OpenAI and Google file-search specifications, including the Vertex AI limitation. Done means the project has a recorded scope decision and, if applicable, a documentation recipe or defined built-in-tool requirements.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai, api
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100