MiMo Multi-Agent DevAssistant

AI-powered multi-agent development pipeline built on Xiaomi MiMo V2.5 models.

A production-grade orchestration platform that coordinates specialized AI agents for automated code development — from planning to code generation, review, testing, and documentation — using the full suite of Xiaomi MiMo V2.5 models.

---

Architecture

┌─────────────────────────────────────────────────────────┐
│                    Orchestrator Agent                   │
│           Coordinates pipeline & manages state          │
└──────┬──────┬──────┬──────┬──────┬──────────────────────┘
       │      │      │      │      │
       ▼      ▼      ▼      ▼      ▼
   ┌───────┐┌──────┐┌───────┐┌──────┐┌──────────┐
   │Planner││Coder ││Review ││Tester││Documenter│
   │ Agent ││Agent ││ Agent ││Agent ││  Agent   │
   └───┬───┘└──┬───┘└───┬───┘└──┬───┘└────┬─────┘
       │       │        │       │         │
       ▼       ▼        ▼       ▼         ▼
   ┌───────────────────────────────────────────────┐
   │           MiMo API Client (OpenAI-compatible) │
   ├───────────────┬───────────────┬───────────────┤
   │ MiMo-V2.5-Pro │   MiMo-V2.5  │ MiMo-V2.5-TTS│
   │  (Reasoning)  │ (Multimodal) │   (Speech)    │
   └───────────────┴───────────────┴───────────────┘

Agent Roles

Agent	Model	Purpose
Planner	MiMo-V2.5-Pro	Decomposes tasks into structured implementation plans
Coder	MiMo-V2.5-Pro	Generates production-ready code from plans
Reviewer	MiMo-V2.5-Pro	Multi-dimensional code review (security, performance, correctness)
Tester	MiMo-V2.5-Pro	Generates comprehensive test suites with pytest
Documenter	MiMo-V2.5	Creates documentation with multimodal understanding
Narrator	MiMo-V2.5-TTS	Voice narration for code reviews and docs

Pipeline Flow

graph LR
    A[Task Input] --> B[Planner Agent]
    B --> C[Coder Agent]
    C --> D[Reviewer Agent]
    D -->|APPROVE| E[Tester Agent]
    D -->|REQUEST_CHANGES| C
    E --> F[Documenter Agent]
    F --> G[Final Output]

Loading

The pipeline includes iterative refinement: if the Reviewer requests changes, the Coder automatically revises the code and re-submits for review, up to a configurable number of iterations.

---

Features

• Multi-Agent Orchestration: Coordinated pipeline with specialized agents that communicate via structured messages
• Iterative Refinement: Automated code improvement loop between Coder and Reviewer agents
• Full MiMo V2.5 Integration: Leverages all three model types:
  • MiMo-V2.5-Pro for deep reasoning and code generation
  • MiMo-V2.5 for multimodal document understanding
  • MiMo-V2.5-TTS for voice narration
• Multiple Interfaces: CLI, REST API (FastAPI), and Web Dashboard (Streamlit)
• Token Usage Tracking: Real-time monitoring of API consumption across all agents
• OpenAI-Compatible: Uses the standard OpenAI SDK for MiMo API integration

---

Quick Start

Installation

# Clone the repository
git clone https://github.com/singgih123/mimo-multi-agent-devassistant.git
cd mimo-multi-agent-devassistant

# Install with pip
pip install -e ".[dev]"

# Set up environment
cp .env.example .env
# Edit .env with your MiMo API key from platform.xiaomimimo.com

Get Your API Key

1. Visit platform.xiaomimimo.com
2. Sign up / Log in
3. Navigate to API Keys
4. Create a new key and add it to your .env file

Usage

CLI

# Run the full multi-agent pipeline
mimo-agent run "Build a REST API for user authentication with JWT tokens" \
  --title "Auth API" --language python

# Review existing code
mimo-agent review src/myapp/auth.py --language python

# Generate tests
mimo-agent test src/myapp/handlers.py

# Create an implementation plan
mimo-agent plan "Design a microservices architecture for an e-commerce platform"

# Check configuration
mimo-agent status

REST API

# Start the API server
uvicorn mimo_agents.api:app --host 0.0.0.0 --port 8000

# Run pipeline via API
curl -X POST http://localhost:8000/pipeline \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Auth Service",
    "description": "Build JWT authentication service",
    "language": "python"
  }'

# Review code via API
curl -X POST http://localhost:8000/review \
  -H "Content-Type: application/json" \
  -d '{
    "code": "def login(user, pwd): return db.query(user, pwd)",
    "language": "python"
  }'

Web Dashboard

streamlit run src/mimo_agents/ui/dashboard.py

---

Configuration

Environment Variable	Default	Description
MIMO_API_KEY	(required)	Your MiMo API key
MIMO_BASE_URL	https://api.xiaomimimo.com/v1	MiMo API endpoint
MIMO_REASONING_MODEL	MiMo-V2.5-Pro	Model for reasoning tasks
MIMO_MULTIMODAL_MODEL	MiMo-V2.5	Model for multimodal tasks
MIMO_TTS_MODEL	MiMo-V2.5-TTS	Model for text-to-speech

---

Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Lint
ruff check src/ tests/

# Type check
mypy src/

---

How It Works

1. Task Intake: User describes a development task via CLI, API, or Web UI
2. Planning Phase: The Planner Agent uses MiMo-V2.5-Pro to decompose the task into structured sub-tasks with technology choices and complexity estimates
3. Code Generation: The Coder Agent generates production-ready code following the plan
4. Review Loop: The Reviewer Agent performs multi-dimensional analysis (correctness, security, performance, maintainability). If changes are needed, the Coder revises automatically
5. Test Generation: The Tester Agent creates a comprehensive pytest test suite
6. Documentation: The Documenter Agent uses MiMo-V2.5's multimodal capabilities to generate Markdown documentation with architecture diagrams
7. Output: All artifacts are collected and presented — code, tests, docs, and usage metrics

---

Tech Stack

• Python 3.10+ — Core language
• OpenAI SDK — MiMo API integration (OpenAI-compatible)
• FastAPI — REST API server
• Streamlit — Web dashboard
• Rich — Terminal UI
• Click — CLI framework
• Pydantic — Data validation
• pytest — Testing

---

License

MIT License — see LICENSE for details.