[feat] Proposal: Increment access_count in ContentsDB when documents are accessed

General
,
1

I have a doubt regarding the knowledge-base implementation, where we maintain two databases:

  • ContentsDB (PostgresDb)

  • VectorDB (pgvector)

In ContentsDB, we have an attribute access_count , and my expectation is that it should be updated every time a document is accessed.

Example

If in a single search I retrieve ~10 documents, then for each unique knowledge content, access_count should be incremented by 1. However, currently in the code this update is not happening which is why I am raising this issue.

Code references fix:

ContentsDB increment method (Postgres)

# libs/agno/agno/db/postgres/postgres.py
def increment_knowledge_access_count(self, id: str) -> Optional[KnowledgeRow]:
    ...
    stmt = (
        update(table)
        .where(table.c.id == id)
        .values(
            access_count=func.coalesce(table.c.access_count, 0) + 1,
            updated_at=int(time.time())
        )
        .returning(*table.c)
    )
    ...

VectorDB search result construction (pgvector)

# libs/agno/agno/vectordb/pgvector/pgvector.py
search_results.append(
    Document(
        id=result.id,
        name=result.name,
        meta_data=result.meta_data,
        content=result.content,
        embedder=self.embedder,
        embedding=result.embedding,
        usage=result.usage,
        content_id=result.content_id  # additional select
    )
)

Knowledge layer search logic

# libs/agno/agno/knowledge/knowledge.py
def search(...):
    ...
    results = self.vector_db.search(query=query, limit=_max_results, filters=filters)
    if self.contents_db:
        unique_content_ids = set(item.content_id for item in results)
        for content_id in unique_content_ids:
            self.contents_db.increment_knowledge_access_count(content_id)
    return results

Proposal

Increment access_count in ContentsDB whenever knowledge content is accessed through a search.

Only increment once per content_id per search (avoid duplicates).

This helps track document usage across both ContentsDB and VectorDB consistently.

Should we proceed with this approach, or is there a better way to handle access_count updates across ContentsDB and VectorDB?

2

Hey @KarthikeyaKollu, Thank you for sharing this request with us.I have forwarded this to our team, we will get back to you after a quick discussion.

Thank you for making Agno better

1 Like
3

can anyone look into this???

4

Hey @KarthikeyaKollu, I have already acknowledged your request above - please check my response.
Is there anything else you are looking in particular?

1 Like