base on A minimal Python framework for building custom AI inference servers with full control over logic, batching, and scaling. <div align='center'> <h1> Build custom inference servers in pure Python <br/> </h1> <h4> Define exactly how inference works for models, agents, RAG, or pipelines. <br/> Control batching, routing, streaming, and orchestration without MLOps glue or config files. </h4> <img alt="Lightning" src="https://pl-bolts-doc-images.s3.us-east-2.amazonaws.com/app-2/ls_banner2.png" width="800px" style="max-width: 100%;"> &nbsp; </div> <div align='center'> <pre> ✅ Custom inference logic ✅ 2× faster than FastAPI ✅ Agents, RAG, pipelines, more ✅ Custom logic + control ✅ Any PyTorch model ✅ Self-host or managed ✅ Multi-GPU autoscaling ✅ Batching + streaming ✅ BYO model or vLLM ✅ No MLOps glue code ✅ Easy setup in Python ✅ Serverless support </pre> <div align='center'> [](https://pepy.tech/projects/litserve) [](https://discord.gg/WajDThKAur)  [](https://codecov.io/gh/Lightning-AI/litserve) [](https://github.com/Lightning-AI/litserve/blob/main/LICENSE) </div> </div> <div align="center"> <div style="text-align: center;"> <a target="_blank" href="#quick-start" style="margin: 0 10px;">Quick start</a> • <a target="_blank" href="#featured-examples" style="margin: 0 10px;">Examples</a> • <a target="_blank" href="#features" style="margin: 0 10px;">Features</a> • <a target="_blank" href="#performance" style="margin: 0 10px;">Performance</a> • <a target="_blank" href="#host-anywhere" style="margin: 0 10px;">Hosting</a> • <a target="_blank" href="https://lightning.ai/docs/litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme" style="margin: 0 10px;">Docs</a> </div> </div> &nbsp; <div align="center"> <a target="_blank" href="https://lightning.ai/docs/litserve/home/get-started?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme"> <img src="https://pl-bolts-doc-images.s3.us-east-2.amazonaws.com/app-2/get-started-badge.svg" height="36px" alt="Get started"/> </a> </div> &nbsp; # Why LitServe? Most serving tools (vLLM, etc..) are built for a single model type and enforce rigid abstractions. They work well until you need custom logic, multiple models, agents, or non standard pipelines. LitServe lets you write your own inference engine in Python. You define how requests are handled, how models are loaded, how batching and routing work, and how outputs are produced. LitServe handles performance, concurrency, scaling, and deployment. Use LitServe to build inference APIs, agents, chatbots, RAG systems, MCP servers, or multi model pipelines. Run it locally, self host anywhere, or deploy with one click on [Lightning AI](https://lightning.ai/litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme). &nbsp; # Want the easiest way to host inference? Over 380,000 developers use [Lightning Cloud](https://lightning.ai/?utm_source=ptl_readme&utm_medium=referral&utm_campaign=ptl_readme), the simplest way to run LitServe without managing infrastructure. Deploy with one command, get autoscaling GPUs, monitoring, and a free tier. No cloud setup required. Or self host anywhere. # Quick start Install LitServe via pip ([more options](https://lightning.ai/docs/litserve/home/install)): ```bash pip install litserve ``` [Example 1](#inference-engine-example): Toy inference pipeline with multiple models. [Example 2](#agent-example): Minimal agent to fetch the news (with OpenAI API). ([Advanced examples](#featured-examples)): ### Inference engine example ```python import litserve as ls # define the api to include any number of models, dbs, etc... class InferenceEngine(ls.LitAPI): def setup(self, device): self.text_model = lambda x: x**2 self.vision_model = lambda x: x**3 def predict(self, request): x = request["input"] # perform calculations using both models a = self.text_model(x) b = self.vision_model(x) c = a + b return {"output": c} if __name__ == "__main__": # 12+ features like batching, streaming, etc... server = ls.LitServer(InferenceEngine(max_batch_size=1), accelerator="auto") server.run(port=8000) ``` Deploy for free to [Lightning cloud](#hosting-options) (or self host anywhere): ```bash # Deploy for free with autoscaling, monitoring, etc... lightning deploy server.py --cloud # Or run locally (self host anywhere) lightning deploy server.py # python server.py ``` Test the server: Simulate an http request (run this on any terminal): ```bash curl -X POST http://127.0.0.1:8000/predict -H "Content-Type: application/json" -d '{"input": 4.0}' ``` ### Agent example ```python import re, requests, openai import litserve as ls class NewsAgent(ls.LitAPI): def setup(self, device): self.openai_client = openai.OpenAI(api_key="OPENAI_API_KEY") def predict(self, request): website_url = request.get("website_url", "https://text.npr.org/") website_text = re.sub(r'<[^>]+>', ' ', requests.get(website_url).text) # ask the LLM to tell you about the news llm_response = self.openai_client.chat.completions.create( model="gpt-3.5-turbo", messages=[{"role": "user", "content": f"Based on this, what is the latest: {website_text}"}], ) output = llm_response.choices[0].message.content.strip() return {"output": output} if __name__ == "__main__": server = ls.LitServer(NewsAgent()) server.run(port=8000) ``` Test it: ```bash curl -X POST http://127.0.0.1:8000/predict -H "Content-Type: application/json" -d '{"website_url": "https://text.npr.org/"}' ``` &nbsp; # Key benefits A few key benefits: - **Deploy any pipeline or model**: Agents, pipelines, RAG, chatbots, image models, video, speech, text, etc... - **No MLOps glue:** LitAPI lets you build full AI systems (multi-model, agent, RAG) in one place ([more](https://lightning.ai/docs/litserve/api-reference/litapi?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme)). - **Instant setup:** Connect models, DBs, and data in a few lines with `setup()` ([more](https://lightning.ai/docs/litserve/api-reference/litapi?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme#setup)). - **Optimized:** autoscaling, GPU support, and fast inference included ([more](https://lightning.ai/docs/litserve/api-reference/litserver?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme)). - **Deploy anywhere:** self-host or one-click deploy with Lightning ([more](https://lightning.ai/docs/litserve/features/deploy-on-cloud?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme)). - **FastAPI for AI:** Built on FastAPI but optimized for AI - 2× faster with AI-specific multi-worker handling ([more]((#performance))). - **Expert-friendly:** Use vLLM, or build your own with full control over batching, caching, and logic ([more](https://lightning.ai/lightning-ai/studios/deploy-a-private-llama-3-2-rag-api?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme)). > ⚠️ Not a vLLM or Ollama alternative out of the box. LitServe gives you lower-level flexibility to build what they do (and more) if you need it. &nbsp; # Featured examples Here are examples of inference pipelines for common model types and use cases. <pre> <strong>Toy model:</strong> <a target="_blank" href="#define-a-server">Hello world</a> <strong>LLMs:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-llama-3-2-vision-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Llama 3.2</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/openai-fault-tolerant-proxy-server?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">LLM Proxy server</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-ai-agent-with-tool-use?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Agent with tool use</a> <strong>RAG:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-private-llama-3-2-rag-api?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">vLLM RAG (Llama 3.2)</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-private-llama-3-1-rag-api?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">RAG API (LlamaIndex)</a> <strong>NLP:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-any-hugging-face-model-instantly?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Hugging face</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-hugging-face-bert-model?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">BERT</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-text-embedding-api-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Text embedding API</a> <strong>Multimodal:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-open-ai-clip-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">OpenAI Clip</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-multi-modal-llm-with-minicpm?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">MiniCPM</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-phi3-5-vision-api-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Phi-3.5 Vision Instruct</a>, <a target="_blank" href="https://lightning.ai/bhimrajyadav/studios/deploy-and-chat-with-qwen2-vl-using-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Qwen2-VL</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-multi-modal-llm-with-pixtral?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Pixtral</a> <strong>Audio:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-open-ai-s-whisper-model?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Whisper</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-an-music-generation-api-with-meta-s-audio-craft?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">AudioCraft</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-an-audio-generation-api?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">StableAudio</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-noise-cancellation-api-with-deepfilternet?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Noise cancellation (DeepFilterNet)</a> <strong>Vision:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-private-api-for-stable-diffusion-2?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Stable diffusion 2</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-an-image-generation-api-with-auraflow?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">AuraFlow</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-an-image-generation-api-with-flux?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Flux</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-super-resolution-image-api-with-aura-sr?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Image Super Resolution (Aura SR)</a>, <a target="_blank" href="https://lightning.ai/bhimrajyadav/studios/deploy-background-removal-api-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Background Removal</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-controlled-image-generation-api-controlnet?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Control Stable Diffusion (ControlNet)</a> <strong>Speech:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-a-voice-clone-api-coqui-xtts-v2-model?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Text-speech (XTTS V2)</a>, <a target="_blank" href="https://lightning.ai/bhimrajyadav/studios/deploy-a-speech-generation-api-using-parler-tts-powered-by-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Parler-TTS</a> <strong>Classical ML:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-random-forest-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Random forest</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-xgboost-with-litserve?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">XGBoost</a> <strong>Miscellaneous:</strong> <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-an-media-conversion-api-with-ffmpeg?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">Media conversion API (ffmpeg)</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/deploy-both-pytorch-and-tensorflow-in-a-single-api?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">PyTorch + TensorFlow in one API</a>, <a target="_blank" href="https://lightning.ai/lightning-ai/studios/openai-fault-tolerant-proxy-server?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme">LLM proxy server</a> </pre> </pre> [Browse 100+ community-built templates](https://lightning.ai/studios?section=serving&utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) &nbsp; # Host anywhere Self-host with full control, or deploy with [Lightning AI](https://lightning.ai/?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) in seconds with autoscaling, security, and 99.995% uptime. **Free tier included. No setup required. Run on your cloud** ```bash lightning deploy server.py --cloud ``` https://github.com/user-attachments/assets/ff83dab9-0c9f-4453-8dcb-fb9526726344 &nbsp; # Features <div align='center'> | [Feature](https://lightning.ai/docs/litserve/features?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | Self Managed | [Fully Managed on Lightning](https://lightning.ai/deploy?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | |----------------------------------------------------------------------|-----------------------------------|------------------------------------| | Docker-first deployment | ✅ DIY | ✅ One-click deploy | | Cost | ✅ Free (DIY) | ✅ Generous [free tier](https://lightning.ai/pricing?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) with pay as you go | | Full control | ✅ | ✅ | | Use any engine (vLLM, etc.) | ✅ | ✅ vLLM, Ollama, LitServe, etc. | | Own VPC | ✅ (manual setup) | ✅ Connect your own VPC | | [(2x)+ faster than plain FastAPI](#performance) | ✅ | ✅ | | [Bring your own model](https://lightning.ai/docs/litserve/features/full-control?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Build compound systems (1+ models)](https://lightning.ai/docs/litserve/home?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [GPU autoscaling](https://lightning.ai/docs/litserve/features/gpu-inference?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Batching](https://lightning.ai/docs/litserve/features/batching?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Streaming](https://lightning.ai/docs/litserve/features/streaming?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Worker autoscaling](https://lightning.ai/docs/litserve/features/autoscaling?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Serve all models: (LLMs, vision, etc.)](https://lightning.ai/docs/litserve/examples?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Supports PyTorch, JAX, TF, etc...](https://lightning.ai/docs/litserve/features/full-control?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [OpenAPI compliant](https://www.openapis.org/) | ✅ | ✅ | | [Open AI compatibility](https://lightning.ai/docs/litserve/features/open-ai-spec?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [MCP server support](https://lightning.ai/docs/litserve/features/mcp?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Asynchronous](https://lightning.ai/docs/litserve/features/async-concurrency?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ✅ | ✅ | | [Authentication](https://lightning.ai/docs/litserve/features/authentication?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | ❌ DIY | ✅ Token, password, custom | | GPUs | ❌ DIY | ✅ 8+ GPU types, H100s from $1.75 | | Load balancing | ❌ | ✅ Built-in | | Scale to zero (serverless) | ❌ | ✅ No machine runs when idle | | Autoscale up on demand | ❌ | ✅ Auto scale up/down | | Multi-node inference | ❌ | ✅ Distribute across nodes | | Use AWS/GCP credits | ❌ | ✅ Use existing cloud commits | | Versioning | ❌ | ✅ Make and roll back releases | | Enterprise-grade uptime (99.95%) | ❌ | ✅ SLA-backed | | SOC2 / HIPAA compliance | ❌ | ✅ Certified & secure | | Observability | ❌ | ✅ Built-in, connect 3rd party tools| | CI/CD ready | ❌ | ✅ Lightning SDK | | 24/7 enterprise support | ❌ | ✅ Dedicated support | | Cost controls & audit logs | ❌ | ✅ Budgets, breakdowns, logs | | Debug on GPUs | ❌ | ✅ Studio integration | | [20+ features](https://lightning.ai/docs/litserve/features?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) | - | - | </div> &nbsp; # Performance LitServe is designed for AI workloads. Specialized multi-worker handling delivers a minimum **2x speedup over FastAPI**. Additional features like batching and GPU autoscaling can drive performance well beyond 2x, scaling efficiently to handle more simultaneous requests than FastAPI and TorchServe. Reproduce the full benchmarks [here](https://lightning.ai/docs/litserve/home/benchmarks?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) (higher is better). <div align="center"> <img alt="LitServe" src="https://pl-bolts-doc-images.s3.us-east-2.amazonaws.com/app-2/ls_charts_v6.png" width="1000px" style="max-width: 100%;"> </div> These results are for image and text classification ML tasks. The performance relationships hold for other ML tasks (embedding, LLM serving, audio, segmentation, object detection, summarization etc...). ***💡 Note on LLM serving:*** For high-performance LLM serving (like Ollama/vLLM), integrate [vLLM with LitServe](https://lightning.ai/lightning-ai/studios/deploy-a-private-llama-3-2-rag-api?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme), use [LitGPT](https://github.com/Lightning-AI/litgpt?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme#deploy-an-llm), or build your custom vLLM-like server with LitServe. Optimizations like kv-caching, which can be done with LitServe, are needed to maximize LLM performance. &nbsp; # Community LitServe is a [community project accepting contributions](https://lightning.ai/docs/litserve/community?utm_source=litserve_readme&utm_medium=referral&utm_campaign=litserve_readme) - Let's make the world's most advanced AI inference engine. 💬 [Get help on Discord](https://discord.com/invite/XncpTy7DSt) 📋 [License: Apache 2.0](https://github.com/Lightning-AI/litserve/blob/main/LICENSE) ", Assign "at most 3 tags" to the expected json: {"id":"11569","tags":[]} "only from the tags list I provide: []" returns me the "expected json"