[Feature] Dakera integration — persistent decay-weighted user memory across RAG chat sessions

Problem

Kotaemon does excellent RAG — users upload documents, ask questions, get grounded answers. But the user's own context (preferences, past questions, learned facts) isn't remembered across chat sessions.

If I tell kotaemon in session 1 that I'm a cardiologist reviewing drug trial data, session 2 starts with no context. The document knowledge is persistent; the user knowledge is not.

Proposed: Dakera as a user memory layer

Dakera is a self-hosted vector memory server with decay weighting. It complements kotaemon's document RAG with user memory — what the user has told the system.

Integration point in libs/ktem/ktem/reasoning/ or the pipeline construction:

from dakera import DakeraClient

_memory = DakeraClient(base_url="http://localhost:3300", api_key="demo")

def augment_with_user_memory(user_id: str, query: str, doc_context: str) -> str:
    """Add user memory context to the retrieved document chunks."""
    resp = _memory.recall(agent_id=user_id, query=query, top_k=3)
    user_ctx = "\n".join(f"- {m.content}" for m in (resp.memories or []))
    if not user_ctx:
        return doc_context
    return f"User context (prior sessions):\n{user_ctx}\n\nDocument context:\n{doc_context}"

def store_exchange(user_id: str, question: str, answer: str) -> None:
    _memory.store_memory(
        agent_id=user_id,
        content=f"Q: {question}\nA: {answer}",
    )

Why this matters

Kotaemon users often build document-centric knowledge bases for specific domains. Domain experts who use kotaemon repeatedly would benefit from the assistant remembering their expertise level, vocabulary, and prior queries — so follow-up questions can skip re-establishing context.

Decay weighting means old preferences naturally fade, keeping the recalled context fresh and relevant.

Setup

docker run -d -p 3300:3300 -e DAKERA_API_KEY=demo ghcr.io/dakera-ai/dakera:latest
pip install dakera

Self-hosted — fits kotaemon's privacy-first model. Happy to prototype as a PR.