Repository navigation
[Enhancement]: Support custom reranker provider with configurable URL and model name #12119
Description
What features would you like to see added?
Add support for a custom reranker type in the web search configuration, allowing users to specify their own Jina-compatible reranking endpoint with a custom URL, API key, and model name.
Currently only jina and cohere are supported as reranker providers, and model names are hardcoded. Many providers (Scaleway, vLLM, LiteLLM, etc.) offer Jina-compatible /v1/rerank endpoints with different model names. The documentation already mentions "Planned Rerankers: Additional third-party reranking services".
More details
The configuration would look like this in librechat.yaml:
webSearch:
rerankerType: "custom"
customRerankerApiUrl: "${CUSTOM_RERANKER_API_URL}"
customRerankerApiKey: "${CUSTOM_RERANKER_API_KEY}"
customRerankerModel: "qwen3-embedding-8b"With corresponding environment variables:
CUSTOM_RERANKER_API_URL=https://api.scaleway.ai/v1/rerank
CUSTOM_RERANKER_API_KEY=your-api-key
CUSTOM_RERANKER_MODEL=qwen3-embedding-8bChanges needed:
- Add
'custom'toRerankerTypeinpackages/data-schemas/src/app/web.ts - Add
customRerankerApiUrl,customRerankerApiKey,customRerankerModelto the web search auth schema - Pass the new config through
handleTools.jstocreateSearchTool() - Depends on companion issue in
@librechat/agents: [Enhancement]: Support custom reranker provider with configurable URL and model name agents#65
Which components are impacted by your request?
Endpoints
Pictures
No response
Related issues
- feat: Add COHERE_API_URL environment variable support for custom Cohere endpoints #9544
- [Enhancement]: Support Custom API endpoint for Cohere ReRanking #8710
- [Bug]: rerankerType config does not permit all allowed types, and this option is not passed #10022
- 🔗 feat: Custom Jina API URL for Web Search Reranking #9236
Code of Conduct
- I agree to follow this project's Code of Conduct
Activity
danny-avila commentedon Mar 7, 2026
Thanks for the request, on my list to finalize this PR: LibreChat-AI/rag-api#190
mdbook commentedon Jun 19, 2026
+1, ran into this exact thing self-hosting against a local Infinity reranker.
two issues with the jina path right now:
jina-reranker-v2-base-multilingualin@librechat/agents(tools/search/rerankers.ts), no way to override itcustomreranker type, just jina and coherejinaApiUrlalready lets you point somewhere else and that part works great (Infinity serves a jina-style/rerank). the problem is the hardcoded model name, so my reranker has to answer tojina-reranker-v2-base-multilingual. my current hack is aliasing the real model under that name with Infinity's--served-model-name. it works, but it's gross: the served name is now decoupled from the actual model, and anything else hitting that endpoint (a memory service, in my case) has to send the fake name too.rerankerType: custom+customRerankerModelwould just fix it. would love to see it land. this is a UI built around local LLMs and reranking is basically the last thing still pinned to a cloud provider's model name.feat: add custom reranker provider support in web search config