RAG: no capabilities on (api) versions above 0.7.8

[anunomac]

What happened?

So a few weeks ago i deployed an instance of librechat (using docker-compose) and successfully enabled everything i wanted to enable: My models via Azure; WebSearch (using Serper, etc) and RAG capabilities using openai embeddigns via Azure also. This was done with version 0.7.8

Last night i attempted to deploy it again, using the latest librechat version and RAG simply doesnt work.
Dragging a txt, pdf or word file to the window displays the error: "no capabilities to handle that file type":

And using the "upload as text" option just tries to use OCR which is not intended:

there's no mention of OCR on my .env or other yml configs.

Some information about the setup
.env

librechat,yaml
simply added the azureopenai endpoint for each model, let me know if you want more info about this

docker-compose-override.yml
Is just copying the librechat.yaml

How do i know the problem is the version? well i just added the line:
image: ghcr.io/danny-avila/librechat:v0.7.8
to the override and it started accepting files again.

Version Information

ghcr.io/danny-avila/librechat-dev latest 34280f5146b2 2 days ago 1.1GB
ghcr.io/danny-avila/librechat-rag-api-dev latest 10f3c05fd82e 3 days ago 7.87GB
ghcr.io/danny-avila/librechat-rag-api-dev-lite latest 2991e83a1e7b 3 days ago 1.48GB
ghcr.io/danny-avila/librechat latest 3d0aff522ece 6 days ago 1.11GB
ghcr.io/danny-avila/librechat v0.7.8 17cfdb1df083 2 months ago 1.02GB

Steps to Reproduce

1. pull the latest images
2. configure RAG
3. try to upload a file

What browsers are you seeing the problem on?

No response

Relevant log output

Other than the logs i posted on the above screenshot showing the OCR attempt theres not much there. I can post rag_api and vectordb logs to show they are working:


docker logs vectordb

2025-07-09 09:07:21.365 UTC [1] LOG:  starting PostgreSQL 15.4 (Debian 15.4-2.pgdg120+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 12.2.0-14) 12.2.0, 64-bit
2025-07-09 09:07:21.366 UTC [1] LOG:  listening on IPv4 address "0.0.0.0", port 5432
2025-07-09 09:07:21.366 UTC [1] LOG:  listening on IPv6 address "::", port 5432
2025-07-09 09:07:21.385 UTC [1] LOG:  listening on Unix socket "/var/run/postgresql/.s.PGSQL.5432"
2025-07-09 09:07:21.401 UTC [29] LOG:  database system was interrupted; last known up at 2025-07-09 08:50:11 UTC
2025-07-09 09:07:23.033 UTC [29] LOG:  database system was not properly shut down; automatic recovery in progress
2025-07-09 09:07:23.039 UTC [29] LOG:  redo starts at 0/19D08C0
2025-07-09 09:07:23.039 UTC [29] LOG:  invalid record length at 0/19D08F8: wanted 24, got 0
2025-07-09 09:07:23.039 UTC [29] LOG:  redo done at 0/19D08C0 system usage: CPU: user: 0.00 s, system: 0.00 s, elapsed: 0.00 s
2025-07-09 09:07:23.050 UTC [27] LOG:  checkpoint starting: end-of-recovery immediate wait
2025-07-09 09:07:23.092 UTC [27] LOG:  checkpoint complete: wrote 3 buffers (0.0%); 0 WAL file(s) added, 0 removed, 0 recycled; write=0.012 s, sync=0.006 s, total=0.048 s; sync files=2, longest=0.003 s, average=0.003 s; distance=0 kB, estimate=0 kB
2025-07-09 09:07:23.107 UTC [1] LOG:  database system is ready to accept connections
2025-07-09 09:07:24.550 UTC [33] WARNING:  there is already a transaction in progress
2025-07-09 09:07:24.553 UTC [33] WARNING:  there is no transaction in progress

docker logs rag_api
2025-07-09 09:07:24,462 - root - INFO - Initialized embeddings of type: <class 'langchain_openai.embeddings.azure.AzureOpenAIEmbeddings'>
/app/app/services/vector_store/async_pg_vector.py:9: LangChainPendingDeprecationWarning: Please use JSONB instead of JSON for metadata. This change will allow for more efficient querying that involves filtering based on metadata. Please note that filtering operators have been changed when using JSONB metadata to be prefixed with a $ sign to avoid name collisions with columns. If you're using an existing database, you will need to create a db migration for your metadata column to be JSONB and update your queries to use the new operators. 
  super().__init__(*args, **kwargs)
2025-07-09 09:07:24,718 - uvicorn.error - INFO - Started server process [1]
2025-07-09 09:07:24,718 - uvicorn.error - INFO - Waiting for application startup.
2025-07-09 09:07:24,718 - root - INFO - Initialized thread pool with 8 workers (CPU cores: 20)
2025-07-09 09:07:24,929 - uvicorn.error - INFO - Application startup complete.
2025-07-09 09:07:24,929 - uvicorn.error - INFO - Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

Screenshots

No response

This is how the working override file looks:

services:

# USE LIBRECHAT CONFIG FILE
  api:
    volumes:
    - type: bind
      source: ./librechat.yaml
      target: /app/librechat.yaml
    image: ghcr.io/danny-avila/librechat:v0.7.8

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

It works, I use it every day. Make sure file_search is included in the agents config: https://www.librechat.ai/docs/configuration/librechat_yaml/object_structure/agents#capabilities

[anunomac]

So considering the setup i shared above, how would i disable those options? i feel like i've tried everything, and i cant find any info on the docs, can you help?

[darrell-ford]

I am having the same issue. I added the mime-types to the YAML config but it appears the app is returning this response, there is nothing in logging that shows that a model or the RAG-API container is even seeing the files. With 7.9 I am only able to upload image files. On the positive side I can now using models from Bedrock and Gemini, not just OpenAI. But the support file types seems to have issues. I don't have time at the moment to look at the source code, so I am going to revert back to 7.8.

[anunomac]

@darrell-ford use "upload for file search"

[darrell-ford]

@darrell-ford use "upload for file search"

Sorry, I am not clear on what you mean here. I don't see that as an option.

[darrell-ford]

I see now where it should be. I have Agents currently disabled for our users. I updated the YAML to look like the following, but the option is not available when I redeploy.

endpoints:
assistants:
disableBuilder: true # Disable Assistants Builder Interface by setting to true
agents:
capabilities: ["file_search", "tools"]

Currently, under Attach Files I only see 'Upload Image'.