polarisllm 1.3.3

🌟 The Ultimate Multi-Model LLM Runtime Platform - Deploy, manage, and serve 300+ language models with OpenAI-compatible APIs. Built on ms-swift for production-ready performance.

🌟 PolarisLLM Runtime Engine

The Ultimate Multi-Model LLM Runtime Platform

PolarisLLM is a production-ready, high-performance runtime engine that transforms how you deploy and serve Large Language Models. Built on the robust ms-swift framework, it provides seamless OpenAI-compatible APIs while enabling dynamic multi-model serving, making it the perfect solution for developers, researchers, and enterprises who need flexible, scalable LLM infrastructure.

🎯 Why PolarisLLM?

🚀 Turn Any Server Into an LLM Powerhouse

• Deploy multiple models simultaneously on a single machine
• Switch between models without restarts or downtime
• Support for 300+ models including Qwen, Llama, DeepSeek, Mistral, and more

⚡ Production-Ready Performance

• Built on battle-tested ms-swift framework
• Automatic resource management and optimization
• Real-time health monitoring and auto-recovery

🔌 Drop-in OpenAI Compatibility

• Use existing OpenAI client libraries without modification
• Seamless integration with popular frameworks like LangChain, LlamaIndex
• Perfect for migration from proprietary APIs to self-hosted solutions

✨ Key Features

🎛️ Dynamic Model Management

• Hot-swap models without server restarts
• Concurrent serving of multiple models on different ports
• Intelligent resource allocation and memory management
• Auto-scaling based on demand

🔗 Universal Compatibility

• OpenAI API compatible - works with existing tools and libraries
• 300+ supported models from HuggingFace and ModelScope
• Multi-modal support - text, vision, code, and audio models
• Streaming responses for real-time applications

🛠️ Developer Experience

• Rich CLI interface with beautiful status displays
• RESTful admin APIs for programmatic control
• YAML configuration for easy model definitions
• Comprehensive logging and error handling

🏗️ Production Ready

• Docker containerization with GPU support
• Health checks and automatic recovery
• Resource monitoring and performance metrics
• Horizontal scaling support

📦 Installation

🎉 Quick Install (Recommended)

# Install from PyPI - that's it!
pip install polarisllm

# Start the engine
polaris start

🐳 Docker Installation

# Run with Docker
docker run -p 7860:7860 polarisllm/polarisllm

# Or with docker-compose
git clone https://github.com/polarisllm/polarisLLM.git
cd polarisLLM
docker-compose up -d

🛠️ Development Installation

git clone https://github.com/polarisllm/polarisLLM.git
cd polarisLLM
pip install -e .
pip install ms-swift[llm] --upgrade

🚀 Quick Start Guide

Step 1: Install & Start ⚡

# Install PolarisLLM (includes ms-swift dependency)
pip install polarisllm

# Start the runtime engine
polarisllm

Server starts on http://localhost:7860 with beautiful web interface

Step 2: Verify Installation ✅

# Check if server is running
curl http://localhost:7860/health

# View available models
curl http://localhost:7860/v1/models

# Open API documentation
# Visit: http://localhost:7860/docs

Step 3: Load and Use Models 🤖

# First, load a model using ms-swift (use python -m swift if swift command not in PATH)
swift deploy --model_type qwen2_5 --model_id Qwen/Qwen2.5-7B-Instruct
# OR: python -m swift deploy --model_type qwen2_5 --model_id Qwen/Qwen2.5-7B-Instruct

# Then use with OpenAI-compatible API
curl -X POST "http://localhost:7860/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen2.5-7b-instruct", 
    "messages": [{"role": "user", "content": "Hello! How are you?"}]
  }'

Step 4: Use with Your Favorite Tools 🔧

# Works with OpenAI Python client
import openai

client = openai.OpenAI(
    base_url="http://localhost:7860/v1",
    api_key="not-required"  # No API key needed!
)

response = client.chat.completions.create(
    model="qwen2.5-7b-instruct",
    messages=[{"role": "user", "content": "Write a Python function to sort a list"}]
)
print(response.choices[0].message.content)

🎮 Real-World Examples

Example 1: Multi-Model AI Assistant

# Load different specialized models
polaris load qwen2.5-7b-instruct      # General chat
polaris load deepseek-coder-6.7b      # Code generation  
polaris load deepseek-vl-7b-chat      # Vision understanding

# Use different models for different tasks
curl -X POST "http://localhost:7860/v1/chat/completions" \
  -d '{"model": "deepseek-coder-6.7b", "messages": [{"role": "user", "content": "Write a REST API in FastAPI"}]}'

Example 2: LangChain Integration

from langchain.llms import OpenAI
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate

# Connect to your local PolarisLLM
llm = OpenAI(
    openai_api_base="http://localhost:7860/v1",
    openai_api_key="not-required",
    model_name="qwen2.5-7b-instruct"
)

# Use with LangChain as usual
prompt = PromptTemplate(template="Explain {topic} in simple terms")
chain = LLMChain(llm=llm, prompt=prompt)
result = chain.run(topic="machine learning")

Example 3: Batch Processing

import asyncio
import aiohttp

async def process_documents():
    models = ["qwen2.5-7b-instruct", "deepseek-coder-6.7b"]
    
    for model in models:
        async with aiohttp.ClientSession() as session:
            async with session.post(
                "http://localhost:7860/v1/chat/completions",
                json={
                    "model": model,
                    "messages": [{"role": "user", "content": "Analyze this document..."}]
                }
            ) as response:
                result = await response.json()
                print(f"{model}: {result}")

🛠️ Available Commands

Server Commands

# Start the runtime engine (main command)
polarisllm

# Alternative start commands
polaris-llm
polaris-server

# Start with custom options
polarisllm --host 0.0.0.0 --port 8080

# View help
polarisllm --help

Model Management (via ms-swift)

# List available models
swift list-models
# OR: python -m swift list-models

# Deploy a chat model
swift deploy --model_type qwen2_5 --model_id Qwen/Qwen2.5-7B-Instruct
# OR: python -m swift deploy --model_type qwen2_5 --model_id Qwen/Qwen2.5-7B-Instruct

# Deploy a vision model  
swift deploy --model_type deepseek_vl --model_id deepseek-ai/deepseek-vl-7b-chat

# Deploy a code model
swift deploy --model_type deepseek --model_id deepseek-ai/deepseek-coder-6.7b-instruct

# Check deployment status
swift list
# OR: python -m swift list

🤖 Supported Models (300+)

PolarisLLM supports the entire ms-swift model ecosystem. Here are some popular choices:

🎯 General Chat Models

• Qwen2.5-7B-Instruct: Alibaba's flagship model - excellent for general tasks
• Llama-3.1-8B-Instruct: Meta's latest - great reasoning capabilities
• Mistral-7B-Instruct: Efficient and fast - perfect for production
• DeepSeek-V2.5: Advanced reasoning and long context support

💻 Code Generation Models

• DeepSeek-Coder-6.7B: State-of-the-art code generation
• CodeQwen1.5-7B: Multi-language programming support
• Qwen2.5-Coder-7B: Latest coding model with enhanced capabilities

👁️ Vision-Language Models

• DeepSeek-VL-7B-Chat: Advanced vision understanding
• Qwen2-VL-7B-Instruct: Multi-modal reasoning
• LLaVA-NeXT: Image analysis and description

🎵 Multi-Modal Models

• Qwen2-Audio: Speech and audio understanding
• Qwen2.5-Omni: Text, image, and audio in one model

See the complete list of 300+ supported models in our Model Catalog

🔧 Configuration

Runtime Configuration (config/runtime.yaml)

host: "0.0.0.0"
port_range_start: 8000
port_range_end: 8100
max_concurrent_models: 5
model_timeout: 300
env_vars:
  CUDA_VISIBLE_DEVICES: "0"
  HF_HUB_CACHE: "./cache/huggingface"

Model Configuration (config/models/*.yaml)

name: "custom-model"
model_id: "path/to/model"
model_type: "qwen2_5"
template: "qwen2_5"
description: "Custom model description"
tags: ["chat", "custom"]
swift_args:
  max_length: 8192
  temperature: 0.7

🌐 API Endpoints

OpenAI Compatible Endpoints

• POST /v1/chat/completions - Create chat completion
• GET /v1/models - List available models

Admin Endpoints

• POST /admin/models/load - Load a model
• POST /admin/models/{model_name}/unload - Unload a model
• GET /admin/models/{model_name}/status - Get model status
• GET /admin/status - Get runtime status
• GET /admin/models/available - List available model configurations
• GET /admin/models/running - List running models

Utility Endpoints

• GET /health - Health check
• GET / - API information

🐳 Docker Deployment

Basic Deployment

docker-compose up -d

With GPU Support

1. Install nvidia-docker2
2. Uncomment GPU section in docker-compose.yml
3. Start with GPU access:

docker-compose up -d

With Redis Cache

docker-compose --profile with-cache up -d

📊 Monitoring

Health Checks

The runtime includes built-in health monitoring:

curl http://localhost:7860/health

Resource Monitoring

View real-time resource usage:

python cli.py status

Logs

# Local logs
tail -f polaris.log

# Docker logs
docker-compose logs -f polaris-runtime

🔌 Integration Examples

Python Client

import openai

client = openai.OpenAI(
    base_url="http://localhost:7860/v1",
    api_key="not-required"
)

response = client.chat.completions.create(
    model="deepseek-vl-7b-chat",
    messages=[
        {"role": "user", "content": "Hello!"}
    ]
)

print(response.choices[0].message.content)

JavaScript Client

import OpenAI from 'openai';

const client = new OpenAI({
    baseURL: 'http://localhost:7860/v1',
    apiKey: 'not-required'
});

const completion = await client.chat.completions.create({
    model: 'qwen2.5-7b-instruct',
    messages: [
        { role: 'user', content: 'Hello!' }
    ]
});

console.log(completion.choices[0].message.content);

🛡️ Production Considerations

1. Resource Management: Monitor GPU/CPU usage and memory consumption
2. Load Balancing: Use reverse proxy for multiple runtime instances
3. Security: Add authentication for admin endpoints
4. Logging: Configure structured logging for production monitoring
5. Scaling: Use Kubernetes for large-scale deployments

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🌟 Use Cases

🏢 Enterprise & Startups

• Private AI Infrastructure: Keep models in-house for data privacy
• Cost Optimization: Reduce API costs by 90% compared to cloud providers
• Multi-tenant Applications: Serve different models to different customers
• A/B Testing: Compare model performance with easy switching

👨‍💻 Developers & Researchers

• Local Development: Test AI features without API costs
• Model Comparison: Evaluate different models on the same dataset
• Fine-tuning Pipeline: Deploy custom fine-tuned models
• Prototype Rapidly: Build AI applications with zero setup friction

📚 Educational & Training

• AI Courses: Provide students with hands-on LLM experience
• Research Projects: Access to latest models for academic research
• Hackathons: Quick setup for AI-focused competitions

🏆 Why Choose PolarisLLM Over Alternatives?

Feature	PolarisLLM	Ollama	text-generation-webui	OpenAI API
Multi-model serving	✅ Concurrent	⚠️ Sequential	❌ Single	✅ Multiple
OpenAI compatibility	✅ Full	❌ Limited	❌ None	✅ Native
Model variety	✅ 300+ models	⚠️ GGUF only	⚠️ Limited	⚠️ Proprietary
Production ready	✅ Yes	⚠️ Basic	❌ No	✅ Yes
Self-hosted	✅ Yes	✅ Yes	✅ Yes	❌ No
Cost	✅ Free	✅ Free	✅ Free	💰 Expensive
Setup time	✅ < 2 minutes	⚠️ 5-10 min	❌ 30+ min	✅ Instant

🤝 Community & Support

Getting Help

• 📖 Documentation: Comprehensive guides and API reference
• 💬 GitHub Discussions: Community Q&A and feature requests
• 🐛 Issue Tracking: Bug reports and feature requests
• 📧 Email Support: [e-mail]

Contributing

• 🍴 Fork & PR: Contributions welcome!
• 🧪 Testing: Help test new models and features
• 📝 Documentation: Improve guides and examples
• 🌍 Translation: Help localize for global users

Stay Updated

• ⭐ Star us on GitHub: Get notifications for releases
• 🐦 Follow @PolarisLLM: Latest updates and tips
• 📰 Newsletter: Monthly model updates and tutorials

🔄 Architecture

┌─────────────────┐    ┌──────────────────┐    ┌─────────────────┐
│   CLI Client    │    │   FastAPI Server │    │  Runtime Core   │
│                 │────│                  │────│                 │
│ - Model Mgmt    │    │ - OpenAI API     │    │ - Model Manager │
│ - Status Check  │    │ - Admin API      │    │ - Process Mgmt  │
└─────────────────┘    └──────────────────┘    └─────────────────┘
                                │                        │
                                │                        │
                       ┌────────▼────────┐    ┌─────────▼─────────┐
                       │  Model Instance │    │  Model Instance   │
                       │                 │    │                   │
                       │ - ms-swift      │    │ - ms-swift        │
                       │ - Port 8000     │    │ - Port 8001       │
                       └─────────────────┘    └───────────────────┘

🎉 Acknowledgments

• Built on the excellent ms-swift framework
• Inspired by OpenAI's API design
• Thanks to the open-source LLM community