Support for index in Store

[sathyarr]

To do store.search(), the embedding is needed.

For that, need to pass embedding configs here,
https://github.com/langchain-ai/langgraph/blob/d73902ae76f106525ef33aa8bfd66b65a93496b3/libs/checkpoint-postgres/langgraph/store/postgres/aio.py#L184

[JoshuaC215]

Sounds good, I would welcome a PR for this. Looks like it's not too complex. It should be optional and driven by config.

[sathyarr]

yeah, Let me raise a PR then, thanks

[sathyarr]

I need help in #261.

How do I pass the IndexConfig in service.py?

I cannot pass it through the .env file. Because, the config basically refers to an embedding_function(),

embeddings = init_embeddings("openai:text-embedding-3-small") # an embedding function

index={
    "embed": embeddings,
    "dims": 1536,
}

should I keep a placeholder in service.py like,
store_index_config: Optional[IndexConfig] = None

and pass it to initialize_store() as initialize_store(store_index_config) ?

---

btw, I see sqlite and postgres are possible as a store in agent-service-toolkit.

• sqlite accepts IndexConfig
• postgres can also accept PostgresIndexConfig which is a sub-class of IndexConfig

I've already mentioned it in commit message.
Is there any other place should I mention?