How to set the graphRAG with local ollama

[edenbuaa]

Description

How to set the graphRAG with local ollama?

Reproduction steps

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Screenshots

![DESCRIPTION](LINK.png)

Logs

No response

Browsers

No response

OS

No response

Additional information

No response

[mkhludnev]

I briefly look through code noticed explicit reference to OpenAI eg https://github.com/Cinnamon/kotaemon/blob/8be8a4a9d048d551aaf79b98e235ea41ec94b695/libs/ktem/ktem/index/file/graph/pipelines.py#L232, it might not be ready for local llms

[mkhludnev]

Another issue is that graphrag is a separate program with own configuration. Which could be configured to other llm but it's a separate question https://github.com/microsoft/graphrag/issues/657

[mkhludnev]

I briefly look through code noticed explicit reference to OpenAI eg

https://github.com/Cinnamon/kotaemon/blob/8be8a4a9d048d551aaf79b98e235ea41ec94b695/libs/ktem/ktem/index/file/graph/pipelines.py#L232 , it might not be ready for local llms

I stuck here

Entity count: 318
2024-09-09 19:15:38.962 - openai._base_client/DEBUG - Sending HTTP Request: POST https://api.openai.com/v1/embeddings

Error embedding chunk {'OpenAIEmbedding': "Error code: 

I think it should pick default configured embeddings instead.

[edenbuaa]

@mkhludnev thanks. I can build the graphRAG and run the pipeline successfully with my local ollama after a investigation about graphrag component.

But I got error which point to the mismatch of embedding between openai and ollama during the query stage. This error is hidden in the kotaemon pipeline, seems a lot of work to fix it.

[kksasa]

if someone can run ollama for graphrag. Please share the steps to us :）

[edenbuaa]

Finally, it works. I will make a PR to this repo

[taprosoft]

Please check the latest version and instruction for setting up GraphRAG here https://github.com/Cinnamon/kotaemon#setup-graphrag.