[Bug]: pg-vector vector store has no proper support

[sujal011]

Check for existing issues

• I have searched the existing issues and checked that my issue is not a duplicate.

What happened?

INFO:     192.168.16.8:59878 - "GET /organization/list HTTP/1.1" 200 OK
INFO:     192.168.16.8:59892 - "GET /team/list HTTP/1.1" 200 OK
06:42:21 - LiteLLM Proxy:ERROR: endpoints.py:435 - RAG Ingest failed: litellm.APIConnectionError: Provider 'pg_vector' is not supported for RAG ingestion. Supported providers: openai, bedrock, gemini, s3_vectors, vertex_ai
Traceback (most recent call last):
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 179, in aingest
    response = await init_response
               ^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 106, in _execute_ingest_pipeline
    ingestion_class = get_ingestion_class(provider)
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 74, in get_ingestion_class

Steps to Reproduce

1. clone and run this openai compatible vector store api server
   https://github.com/[BerriAI/litellm-pgvector](https://github.com/BerriAI/litellm-pgvector)
2. create vector store in the litellm proxy ui

3. when clicking on the creat vector stire button you will get this error-

INFO:     192.168.16.8:59878 - "GET /organization/list HTTP/1.1" 200 OK
INFO:     192.168.16.8:59892 - "GET /team/list HTTP/1.1" 200 OK
06:42:21 - LiteLLM Proxy:ERROR: endpoints.py:435 - RAG Ingest failed: litellm.APIConnectionError: Provider 'pg_vector' is not supported for RAG ingestion. Supported providers: openai, bedrock, gemini, s3_vectors, vertex_ai
Traceback (most recent call last):
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 179, in aingest
    response = await init_response
               ^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 106, in _execute_ingest_pipeline
    ingestion_class = get_ingestion_class(provider)
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 74, in get_ingestion_class

Relevant log output

INFO:     192.168.16.8:59878 - "GET /organization/list HTTP/1.1" 200 OK
INFO:     192.168.16.8:59892 - "GET /team/list HTTP/1.1" 200 OK
06:42:21 - LiteLLM Proxy:ERROR: endpoints.py:435 - RAG Ingest failed: litellm.APIConnectionError: Provider 'pg_vector' is not supported for RAG ingestion. Supported providers: openai, bedrock, gemini, s3_vectors, vertex_ai
Traceback (most recent call last):
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 179, in aingest
    response = await init_response
               ^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 106, in _execute_ingest_pipeline
    ingestion_class = get_ingestion_class(provider)
  File "/usr/lib/python3.13/site-packages/litellm/rag/[main.py](http://main.py)", line 74, in get_ingestion_class

What part of LiteLLM is this about?

UI Dashboard

What LiteLLM version are you on ?

v1.82.3

Twitter / LinkedIn details

No response

[sujal011]

Could you please assign this bug to me so I can work on it?

[chim1aap]

I ran into the same issue today as well.

[maycuatroi1]

Hit the same wall from the Milvus side - a self-hosted vector store rejected at RAG ingest with:

Provider 'milvus' is not supported for RAG ingestion. Supported providers: openai, bedrock, gemini, s3_vectors, vertex_ai

Same root cause as this issue: litellm/rag/main.py -> get_ingestion_class -> INGESTION_REGISTRY only knows the managed providers, so any self-hosted store (pg_vector here, Milvus for me) gets rejected.

I opened #30388 which extends that registry with a milvus ingestion provider (REST API v2, no pymilvus dep, auto-creates the collection). The same pattern - adding a BaseRAGIngestion subclass + one INGESTION_REGISTRY entry - is what pg_vector would need here.

[flefevre]

Any idea why the PR request has been closed?