add milvus support

[ArnavAgrawal03]

No description provided.

[jazzberry-ai]

Bug Report

Name	Severity	Example test case	Description
Missing Milvus URI/API Key validation in core/services_init.py	Medium	Set VECTOR_STORE_PROVIDER to "milvus", but do not define MILVUS_URI in the environment. Run the application.	When VECTOR_STORE_PROVIDER is set to "milvus", the code does not validate the presence of the MILVUS_URI and MILVUS_API_KEY environment variables in core/services_init.py. This will cause a runtime exception when MilvusClient is initialized without the required credentials. A similar check as done for POSTGRES_URI should be added for Milvus.
Invalid Milvus collection schema in core/vector_store/milvus_multivector_store.py	High	Set MULTIVECTOR_PROVIDER to "milvus". Run the application and attempt to ingest a document.	The _create_collection method in core/vector_store/milvus_multivector_store.py creates a Milvus collection schema where both the collection itself (auto_id=True in create_schema) and the "pk" field have auto_id=True. Milvus only allows one auto-id field per collection. This will cause the collection creation to fail, preventing the application from storing multi-vector embeddings in Milvus.
Missing batch size validation when data_rows is empty in MilvusMultiVectorStore	Low	Configure a large MILVUS_BATCH_SIZE in the settings. Attempt to ingest an empty document, resulting in an empty data_rows list.	In MilvusMultiVectorStore.store_embeddings, the safe_batch_size is calculated dynamically using _estimate_batch_size based on a sample row. However, if the data_rows list is empty (e.g., due to an empty document or filtering), the _estimate_batch_size function is never called. The code then uses the configured batch_size which may be too large for the Milvus instance or even be zero. It should add a check to ensure safe_batch_size is at least 1 in this scenario.

_{Comments? Email us.}