Skip to content

Feature: Persistent cross-project memory via Dakera — agents recall prior related work #646

Description

@ferhimedamine

ChatDev 2.0 is an excellent multi-agent platform, but agents re-derive the same architectural decisions from scratch on every new task. This proposes Dakera (https://dakera.ai) as a persistent memory layer so agents can recall patterns from prior related projects.

Problem: When users give ChatDev similar tasks repeatedly, agents re-derive the same architectural decisions every time. A PM agent doesn't remember that last month's REST API project used FastAPI; an engineer doesn't recall the testing patterns used in a similar project.

Proposed Integration:
At ChatChain start, recall(task_description) retrieves memories from similar prior projects. At ChatChain end, store_memory(project_summary) persists key decisions.

In chat_chain.py:

  • Before chain execution: inject recalled memories into initial system context
  • After chain execution: store a summary of decisions made

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

Why Dakera fits ChatDev's philosophy: Both are fully self-hosted. Dakera's decay weighting means more recent successful project patterns surface preferentially — exactly what a simulated software company would want.

Happy to open a PR.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions