use-local-llm

React hooks for streaming responses from local LLMs — Ollama, LM Studio, llama.cpp, and any OpenAI-compatible endpoint. Zero server required. Browser → localhost, directly.

The Problem

Vercel AI SDK is the standard for AI in React — but it requires server routes. Its React hooks (useChat, useCompletion) POST to your API routes, which then call the LLM. This architecture makes it impossible to call http://localhost:11434 directly from the browser.

If you're prototyping with Ollama, LM Studio, or llama.cpp, you don't need a server in between. You need one hook that talks directly to your local model.

The Solution

use-local-llm gives you:

• Direct browser → localhost streaming — no server, no API routes
• Multi-backend support — Ollama, LM Studio, llama.cpp, any OpenAI-compatible endpoint
• Full chat state management — message history, abort, clear, error handling
• Token-by-token streaming — real-time text rendering with onToken callbacks
• Zero runtime dependencies — only a peer dependency on React
• 2.8 KB gzipped — smaller than most icons

Comparison with Vercel AI SDK

Feature	use-local-llm	Vercel AI SDK
Browser → localhost	✅ Direct	❌ Needs API routes
Server required	❌ None	✅ Next.js/Express
Ollama support	✅ Native	⚠️ Via server proxy
LM Studio support	✅ Native	⚠️ Via server proxy
llama.cpp support	✅ Native	⚠️ Via server proxy
Bundle size	2.8 KB gzipped	~45 KB gzipped
Runtime dependencies	0	Multiple
Streaming protocols	SSE + NDJSON	SSE only
TypeScript	Strict	Strict
React version	≥17	≥18

When to Use

Use use-local-llm when you:

• Are building prototypes or tools that talk to a local LLM
• Want to skip setting up API routes just to proxy localhost
• Need a lightweight, zero-dependency solution
• Want to support multiple backends (Ollama, LM Studio, llama.cpp)

Use Vercel AI SDK when you:

• Need cloud LLM providers (OpenAI, Anthropic, etc.)
• Have a full Next.js app with API routes already
• Need features like RAG, tool calling, or structured output