vectorstores: add Milvus

[pattonjp]

In quest to run langchaingo locally without network dependencies. I've added the milvus vector store and huggingface's text-embeddings-inference api client.

PR Checklist

• Read the Contributing documentation.
• Read the Code of conduct documentation.
• Name your Pull Request title clearly, concisely, and prefixed with the name of the primarily affected package you changed according to Good commit messages (such as memory: add interfaces for X, Y or util: add whizzbang helpers).
• Check that there isn't already a PR that solves the problem the same way to avoid creating a duplicate.
• Provide a description in this PR that addresses what the PR is solving, or reference the issue that it solves (e.g. Fixes #123).
• Describes the source of new concepts.
• References existing implementations as appropriate.
• Contains test coverage for new functions.
• Passes all golangci-lint checks.

fixes #59

[tmc]

@pattonjp why close this?

[pattonjp]

@tmc created in this repo accidentally. will reopen as draft while testing.

[wangzhuzhen]

@pattonjp @tmc How about this vectorstores Milvus？we need this 😄

[tmc]

Thanks for this contribution! We're in the process of coming up with a pattern to expand support while keeping the direct dependent library list small/minimal.

[tmc]

Discussion happening on #369 but I don't want to hold this up. LGTM

[tmc]

@pattonjp can the text-embeddings-router go into the docker-compose config?

[tmc]

@pattonjp why not use ollama for embeddings? edit: I filed an issue to make another example.

Thoughts on getting langchaingo linked in the https://github.com/milvus-io/milvus readme?

[pattonjp]

@pattonjp can the text-embeddings-router go into the docker-compose config?

sure, i can see that would make a simpler example. i'll submit a new pr. originally i had it running in compose with the cpu but found it to be too slow and adding all possible gpu configurations did not seem worth the complexity.

[pattonjp]

@pattonjp why not use ollama for embeddings? edit: I filed an issue to make another example.

Thoughts on getting langchaingo linked in the https://github.com/milvus-io/milvus readme?

@tmc
the example started as a rag example, scaled it back but looks like i could have cleaned up more (i see you did that already... thanks)

as for getting on the milvus readme, i'd say go for it.

[eliben]

Hi @pattonjp you've added embeddings/tei here; can you PTAL at #379 and #374 ? We're refactoring everything in embeddings to remove code duplication and use some common interfaces and APIs.

Do you think you could refactor embeddings/tei to do the same, perhaps leaving only things specific to this provider, if needed? Happy to review!

[eliben]

Specifically, the tei provider has some additional options - but could these be passed to the underlying client directly, instead of to the embedder? The embedder then just wraps any client

[pattonjp]

@eliben sure will have pr today