[CHORE]: Increase Ollama timeout

[Nathan8489]

How are you running AnythingLLM?

Docker (local)

What happened?

os: linux
version: mintplexlabs/anythingllm@sha256:1d994f027b5519d4bc5e1299892e7d0be1405308f10d0350ecefc8e717d3154f (latest version before lancedb dep bump )

problem from: server/utils/EmbeddingEngines/ollama/index.js

The function embedChunks tries to send all text chunks in one connection. However, if all fetches cannot complete within 5 minutes, something (possibly Node.js) determines that the connection has timed out and terminates it, causing it to fail.

I temporarily resolved this by awaiting each chunk fetch individually (40 seconds per chunk, 8192 tokens).

Are there known steps to reproduce?

• Get a really slow Ollama service (CPU).
• Set the embedding provider to Ollama.
• Try to move a large document to the workspace.
• Wait for the embedding process...
• It failed exactly after 5 minutes.

[mingLvft]

Has the problem been solved？

[timothycarambat]

There is no PR and the issue is open - so no.

[travisgu]

I encountered same issue. The Ollama service returned 500 error at exact 5 mintues because the http request was cancelled. So it can only embedding small PDF files.

Is there any config file I can increase the timeout limit?

[timothycarambat]

Moving conversation to #1585