qwen-memory-agent

A benchmarked, MCP-native persistent-memory agent built on Qwen Cloud (Alibaba Cloud / DashScope). Submitted to the Qwen Cloud Hackathon, Track 1 — MemoryAgent.

The agent itself decides — via Qwen function-calling — when to remember, recall, or forget. It carries user preferences across sessions, forgets superseded facts, and recalls the right memories inside a tight token budget — and proves it with numbers against naive baselines.

Why it's different

Most memory agents are "stuff everything into RAG and hope." This one adds:

• Agentic memory via Qwen function-calling — the model invokes remember / recall / forget tools through a real agent loop. It's an agent with memory, not a database with an LLM bolted on.

• Supersession-aware forgetting — when a new fact contradicts an old one of the same kind, the old record is retired (not just buried under recency).

• Budget-constrained recall — retrieval greedily packs the most useful memories until a configurable token budget is hit, so context stays small and relevant.

• A reproducible benchmark — synthetic multi-session personas, a held-out query set, and baselines (no-memory / full-history / naive-RAG / ours), scored on recall accuracy, staleness rate, and a context-efficiency curve.

Related MCP server: Memory Crystal MCP Server

Architecture

flowchart TB
    U["MCP client / demo UI"]

    subgraph ecs["Alibaba Cloud ECS (Singapore)"]
        API["FastAPI backend<br/>/chat · /health"]
        AGENT["MemoryAgent loop<br/>Qwen function-calling"]
        MCP["FastMCP server<br/>memory.remember / recall / forget / stats"]
        ENG["Memory Engine<br/>write · retrieve · forget<br/>supersession + token-budget packing"]
        QD[("Qdrant<br/>vector store")]
    end

    DS["Qwen Cloud / DashScope-intl<br/>reasoning model + text-embedding"]

    U -->|HTTP| API
    U -.->|MCP| MCP
    API --> AGENT
    AGENT -->|"decides which tool to call"| ENG
    MCP --> ENG
    AGENT <-->|"chat + tool specs"| DS
    ENG <-->|"embed"| DS
    ENG <--> QD

The agent loop (/chat) lets Qwen choose tool calls; the same memory engine is also exposed directly over MCP for any MCP client. The Qwen client has bounded retry/backoff for resilience.

Stack

Python · FastAPI · Qwen function-calling agent loop · FastMCP · openai SDK → DashScope-intl · Qwen text-embedding · Qdrant.

Quickstart

uv sync
cp .env.example .env   # set DASHSCOPE_API_KEY + DASHSCOPE_BASE_URL
uv run pytest -q       # tests run fully mocked — zero Qwen credit spend

Benchmark results

Reproducible and fully offline — uv run python -m benchmark.run uses a deterministic keyword embedder, so the harness measures the memory engine's ranking + supersession logic (not embedding noise) and costs zero Qwen credits.

B3 is the only system that recalls the current preference (1.00) and never surfaces a superseded one (0.00 staleness). B1 and B2 match on recall but re-surface the retired "coffee" preference on half the queries — because neither has a notion of "this fact was replaced." Staleness = fraction of queries whose answer contained a retired fact (lower is better), over the synthetic multi-session persona set in benchmark/generate.py. Supersession-aware forgetting is what separates B3.

License

MIT — see LICENSE.