[Question]: [openapi]/api/v1/retrieval, The metadata_condition field in the request body isn't being used to filter results. Could you please check/confirm?

[jcl-jcl]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

Describe your problem

Version： v0.20.5
[openapi]/api/v1/retrieval, The metadata_condition field in the request body isn't being used to filter results. Could you please check/confirm?

[dosubot]

You are correct: in RAGFlow v0.20.5, the /api/v1/retrieval endpoint does not use the metadata_condition field to filter results. The implementation in api_app.py only processes parameters like kb_id, doc_ids, and question, and does not extract or apply metadata_condition for filtering at any point in the retrieval logic. This is consistent with the retrieval backend (Dealer class), which only supports filtering by kb_ids and doc_ids, not by metadata or tags see code reference.

Metadata-based filtering using metadata_condition is implemented in some SDK/variant endpoints (like /sdk/doc.py and /sdk/dify_retrieval.py), where it is used to filter document IDs before retrieval, but this logic is not present in the main OpenAPI endpoint see SDK example.

There is an open pull request addressing a related metadata retrieval bug, but it does not add metadata_condition filtering to the core endpoint and is not yet merged see PR #9811.

Let me know if you want more details on workarounds or future plans for this feature.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[jcl-jcl]

Compared with #9811
, my issue is that I passed in metadata_condition, but it did not take effect, resulting in too much data being returned.

[dosubot]

Your issue is valid: in RAGFlow v0.20.5, the /api/v1/retrieval endpoint does not use the metadata_condition field to filter results, so passing it in will not affect the output and can lead to excessive data being returned. The fix in PR #9811 is related but does not address this—it's focused on handling cases where documents lack metadata, not on enabling metadata_condition filtering in the main OpenAPI endpoint. As of now, metadata_condition filtering is not supported in /api/v1/retrieval, and this feature is planned for future releases but not yet available see discussion.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[Llala0928]

@jcl-jcl Do you know about this feature?

[jcl-jcl]

@jcl-jcl Do you know about this feature?

As far as I know, it will be used in both the search agent and the Dify SDK.

[JosephFang03]

That is to say, it's not possible to implement metadata filtering via the /api/v1/retrieval method right now?

[yathaarthmagi]

Based on the docs it seems like metadata filtering is supported during retrieval via the api, but on looking into the code, it's actually not.
This is a bummer, since metadata based retrieval filtering seems like a pretty basic requirement in 2025.
We have been evaluating ragflow to migrate from our current implementation for knowledge base management, but without metadata filtering support, or at least some future commitment to add it, this might not work out for us.

[man111222]

May I ask if there is a clear version that will support metadata search?Currently, I am at version 0.21.0.@dosu.