[feat]:Add support & docs for connecting to local OpenAI-compatible endpoints

[Arindam200]

It would be really helpful if Memori supported (and documented) how to connect to local OpenAI-compatible endpoints.

Motivation

Many developers run models locally using projects like llama.cpp, koboldcpp, or jan.ai.

Some users don’t want to send data to external APIs for privacy or cost reasons.

Having a straightforward way to configure Memori with local endpoints would make it more flexible and self-hosting–friendly.

Proposal

Add instructions in the README on how to:

• Point Memori to a local LLM endpoint (OpenAI API-compatible).
• Configure embeddings with a different base URL than the LLM (since some setups split LLMs and embedding services).

Benefits

• Makes Memori usable in fully local / offline setups.
• Supports wider community adoption across different backends.
• Increases flexibility for developers working with custom infra.

[chisleu]

Some people don't use APIs just because they can run models locally.

+1 to this issue. This is a really interesting project, but it's difficult to use for most people.

If you stand up a product similar to LM Studio in terms of ease of use, and configurable, so that it can use LM Studio as a backend, WITH an open AI compatible front end, everyone will use it. literally everyone. After saying that I'm about to vibe code one myself because I think it's such an opportunity.

+1 to the project

[valdecircarvalho]

It would be really helpful if Memori supported (and documented) how to connect to local OpenAI-compatible endpoints.

Motivation

Many developers run models locally using projects like llama.cpp, koboldcpp, or jan.ai.

Some users don’t want to send data to external APIs for privacy or cost reasons.

Having a straightforward way to configure Memori with local endpoints would make it more flexible and self-hosting–friendly.

Proposal

Add instructions in the README on how to:

• Point Memori to a local LLM endpoint (OpenAI API-compatible).
• Configure embeddings with a different base URL than the LLM (since some setups split LLMs and embedding services).

Benefits

• Makes Memori usable in fully local / offline setups.
• Supports wider community adoption across different backends.
• Increases flexibility for developers working with custom infra.

I second that. I understand you are using the OpenAI API to "crunch" the data. But having the option to use another provider/model would be great. I have a product to modernize legacy applications and I need a memory tool to add to it.

[harshalmore31]

@Arindam200 @chisleu @valdecircarvalho Memori now supports connecting to local OpenAI-compatible providers such as Ollama and LM Studio. Our approach does not rely on embeddings or vector search; instead, Memori uses a SQL-first design with structured entity extraction and relationship mapping for transparent, auditable memory.

We’ve just released Memori-v2 with this update. You can upgrade via:

pip install memorisdk --upgrade

Please see the release notes for details.

[Awwtifishal]

Thank you. I would have put some example in a more visible location, such as in the release notes or the readme. I've eventually found examples here and here.

[harshalmore31]

Thanks for the feedback! We’ll make examples more visible in the README and release notes going forward.