Knowledge filters for agentic RAG in PGVector

[ravishrawal]

Hi, I noticed knowledge filters don't work for pgvector stores as yet, even if we adhere to the meta_data structure in the DB. When is this planned to be released and what is the current workaround? Thanks!

[kausmeows]

Hey @ravishrawal can you explain a bit more or share the error? Filters are supported on pgvector, did you refer from here- https://github.com/agno-agi/agno/blob/main/cookbook/agent_concepts/knowledge/filters/filtering_pgvector.py ??

Docs for more context- https://docs.agno.com/filters/introduction

[ravishrawal]

Hey @kausmeows is agentic filtering supported for PGVector? I had heard it isn't (only for qdrant & lanceDB) and we receive an error with no further information when attempting this.

If so, could you guide me on how to use my existing postgres table for this? Knowledge works well and our table follows the expected schema for this, but now I'd want to use the meta_data field for filtering without reloading records individually every time.

vector_db = PgVector(
    table_name="regulatory_document_chunks",  # use the existing table
    db_url=get_db_url(),
    embedder=OpenAIEmbedder(),  # only needed for query‑time embeddings
    auto_upgrade_schema=False,  # keep Agno from altering your table
    reranker=CohereReranker(model="rerank-v3.5"),
    search_type=SearchType.hybrid,
)

knowledge_base = AgentKnowledge(vector_db=vector_db)