hanyeol/model-compose — Portable AI runtime inspired by docker-compose.

Description and use

**Brief Summary**
hanyeol/model‑compose is a portable AI‑runtime that lets you describe agents, RAG pipelines, and MCP servers in a single Docker‑Compose‑style YAML file and launch them anywhere. By turning isolated prompts and tools into declarative, repeatable workflows, it makes multi‑agent orchestration as simple as running docker‑compose up.

**Value**

• **Unified orchestration** – One YAML definition replaces scattered scripts, config files, and ad‑hoc glue code, giving teams a clear, version‑controlled source of truth for complex AI pipelines.
• **Tool‑use and memory standardisation** – Built‑in support for agent memory stores and tool‑integration pipelines means you can reliably reproduce the same “thinking” process across environments.
• **Portability** – Because the runtime is language‑agnostic and can be containerised, the same workflow runs on local dev machines, on‑prem clusters, or any cloud provider without modification.

**Practical Adoption Path**

1. **Prototype** – Clone the repo, write a model-compose.yml that wires together your existing LLM endpoints, vector stores, and custom tools. Run model-compose up locally to validate the end‑to‑end flow.
2. **Integrate** – Replace ad‑hoc orchestration scripts with the YAML definition; use the provided Python SDK or CLI to trigger runs from CI/CD pipelines or higher‑level applications.
3. **Scale** – Deploy the same YAML to a Kubernetes or Docker‑Swarm cluster; the runtime automatically handles service discovery, health‑checks, and resource limits, letting you scale agents horizontally.

**Production Readiness**

• **Activity & community** – The project shows recent commits (last update 2026‑06‑24), 73 GitHub stars, and a modest but engaged fork base, indicating active maintenance.
• **Maturity** – Core features (API/SDK/CLI, language metadata, topic tagging) are already exposed, and the Python implementation aligns with most AI stacks.
• **Risk considerations** – No major licensing or metadata issues have been identified, but a final security audit and confirmation of long‑term maintainers are advisable before a full production rollout.

Overall, model‑compose offers a high‑signal, low‑friction way to bring reproducible, multi‑agent AI workflows into production, with a clear path from local testing to scalable deployment.

Problem

hanyeol/model-compose helps turn isolated prompts and tools into repeatable agent workflows.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• coordinate multi-agent workflows
• add tool-use pipelines
• standardize agent memory

Evidence

Portable AI runtime inspired by docker-compose. Compose agents, RAG pipelines, and MCP servers in one YAML file and run them anywhere.

Source