Skip to content

hotea/LocalAI-Bridge

BranchesTags

Repository files navigation

LocalAI-Bridge

Local AI Gateway — expose locally authenticated AI tools as OpenAI-compatible and Anthropic-compatible HTTP APIs.

Runs on 127.0.0.1 only. Credentials are read from existing local config files — no tokens are extracted, forwarded, or stored.

Quick Start

git clone https://github.com/hotea/LocalAI-Bridge.git
cd LocalAI-Bridge
npm install
npm run dev

Or one-click install:

curl -fsSL https://raw.githubusercontent.com/hotea/LocalAI-Bridge/main/install.sh | bash

The server starts at http://127.0.0.1:6271.

Supported Providers

Provider Method Auth
Claude Code Subprocess (claude CLI) Auto-detect via claude --version
GitHub Copilot HTTP API (api.githubcopilot.com) Auto-scan ~/.config/github-copilot/
Kiro HTTP API (AWS CodeWhisperer) Auto-scan ~/.aws/sso/cache/
Antigravity HTTP API (Google Cloud AI) Auto-scan Antigravity IDE local data, or config.json

Prerequisites: Node.js >= 20 and at least one provider configured locally.

API Endpoints

Method Path Description
GET /health Provider health status
GET /v1/models List available models
POST /v1/chat/completions OpenAI chat completions (streaming & non-streaming)
POST /v1/messages Anthropic Messages API (streaming & non-streaming)

Usage Examples

Health Check & List Models

# Health check
curl http://127.0.0.1:6271/health

# List all available models
curl http://127.0.0.1:6271/v1/models

Claude Code CLI

Non-streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-code:sonnet",
    "messages": [{"role": "user", "content": "Say hello in one sentence"}]
  }'

Streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-code:opus",
    "messages": [{"role": "user", "content": "Write a Python function to reverse a string"}],
    "stream": true
  }'

Available models: claude-code:sonnet, claude-code:opus, claude-code:haiku

GitHub Copilot

Non-streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot:claude-sonnet-4",
    "messages": [{"role": "user", "content": "Explain closures in JavaScript"}]
  }'

Streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot:gpt-4o",
    "messages": [{"role": "user", "content": "Write a binary search algorithm in Go"}],
    "stream": true
  }'

With system prompt:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot:claude-sonnet-4",
    "messages": [
      {"role": "system", "content": "You are a helpful coding assistant. Reply concisely."},
      {"role": "user", "content": "How do I reverse a linked list?"}
    ]
  }'

Available models: copilot:claude-sonnet-4, copilot:claude-3.7-sonnet, copilot:gpt-4o, copilot:gpt-4.1, copilot:o4-mini, copilot:gemini-2.0-flash, copilot:gemini-2.5-pro

Kiro (AWS CodeWhisperer)

Non-streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kiro:claude-sonnet-4",
    "messages": [{"role": "user", "content": "What is the difference between TCP and UDP?"}]
  }'

Streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kiro:claude-sonnet-4.5",
    "messages": [{"role": "user", "content": "Write a Rust function that finds the longest common subsequence"}],
    "stream": true
  }'

Multi-turn conversation:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kiro:claude-sonnet-4",
    "messages": [
      {"role": "user", "content": "What is a binary tree?"},
      {"role": "assistant", "content": "A binary tree is a tree data structure where each node has at most two children."},
      {"role": "user", "content": "How do I traverse one in-order?"}
    ]
  }'

Available models: kiro:claude-sonnet-4, kiro:claude-sonnet-4.5, kiro:claude-3-7-sonnet, kiro:claude-haiku-4.5

Antigravity (Google Cloud AI)

Non-streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "antigravity:claude-sonnet-4-5",
    "messages": [{"role": "user", "content": "Explain quantum computing in simple terms"}]
  }'

Streaming:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "antigravity:gemini-3-pro-high",
    "messages": [{"role": "user", "content": "Write a TypeScript function to debounce API calls"}],
    "stream": true
  }'

With system prompt:

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "antigravity:claude-opus-4-6-thinking",
    "messages": [
      {"role": "system", "content": "You are an expert software architect."},
      {"role": "user", "content": "Design a scalable microservices architecture"}
    ]
  }'

Available models: antigravity:claude-sonnet-4-5, antigravity:claude-sonnet-4-5-thinking, antigravity:claude-opus-4-5-thinking, antigravity:claude-opus-4-6-thinking, antigravity:gemini-3-pro-high, antigravity:gemini-3-pro-low, antigravity:gemini-3-flash

Anthropic Messages API

All providers support the Anthropic Messages API format:

curl -X POST http://127.0.0.1:6271/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: not-needed" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "copilot:claude-sonnet-4",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Streaming:

curl -X POST http://127.0.0.1:6271/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: not-needed" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "antigravity:claude-sonnet-4-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

OpenAI SDK

Python:

from openai import OpenAI

client = OpenAI(base_url="http://127.0.0.1:6271/v1", api_key="not-needed")

# Non-streaming
response = client.chat.completions.create(
    model="copilot:claude-sonnet-4",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

# Streaming
stream = client.chat.completions.create(
    model="copilot:claude-sonnet-4",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
)
for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="")

Node.js:

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "http://127.0.0.1:6271/v1", apiKey: "not-needed" });

const stream = await client.chat.completions.create({
  model: "copilot:gpt-4o",
  messages: [{ role: "user", content: "Hello!" }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

Anthropic SDK

Python:

from anthropic import Anthropic

client = Anthropic(base_url="http://127.0.0.1:6271", api_key="not-needed")

message = client.messages.create(
    model="copilot:claude-sonnet-4",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)
print(message.content[0].text)

Node.js:

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({ baseURL: "http://127.0.0.1:6271", apiKey: "not-needed" });

const message = await client.messages.create({
  model: "copilot:claude-sonnet-4",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(message.content[0].text);

Model Naming

Use the provider:model format:

Provider Example Models
claude-code claude-code:sonnet, claude-code:opus, claude-code:haiku
copilot copilot:claude-sonnet-4, copilot:gpt-4o, copilot:gemini-2.5-pro
kiro kiro:claude-sonnet-4, kiro:claude-sonnet-4.5
antigravity antigravity:claude-sonnet-4-5, antigravity:gemini-3-pro-high

Built-in aliases: gpt-4 / gpt-4o → copilot:gpt-4o, gpt-4o-mini / gpt-3.5-turbo → copilot:o4-mini

Antigravity Setup

If the Antigravity IDE is installed locally, credentials are auto-scanned from its data directory — no configuration needed.

If auto-scan doesn't find credentials, you can provide them manually:

Option 1 — OAuth login flow:

  1. Start the server: npm run dev
  2. curl -X POST http://127.0.0.1:6271/auth/antigravity/login — returns a Google OAuth URL
  3. Open the URL in a browser, sign in with your Google account
  4. After redirect, the refresh token is saved to config.json automatically
  5. Restart the server

Option 2 — Manual configuration:

Add to config.json:

{
  "providers": {
    "antigravity": {
      "refreshToken": "1//your-refresh-token-here"
    }
  }
}

Configuration

Copy config.default.json to config.json to override settings (gitignored):

cp config.default.json config.json

Environment variable overrides:

Variable Description
LOCALAI_HOST Bind address (default: 127.0.0.1)
LOCALAI_PORT Port (default: 6271)
LOCALAI_CLAUDE_BINARY Path to claude CLI binary

Security

  • Server binds to 127.0.0.1 only (hostname binding + middleware double-check)
  • No tokens are extracted, forwarded, or stored by this service
  • Copilot: reads the local OAuth token and exchanges it for a short-lived session token via GitHub's official API
  • Kiro: reads the local access/refresh token and refreshes it via the official Kiro/AWS auth endpoints
  • Claude Code: calls the CLI as a subprocess — inherits the user's existing authentication
  • Antigravity: uses a user-provided Google OAuth refresh token, auto-refreshes access tokens
  • spawn() with argument arrays (no shell injection)

License

MIT

LocalAI-Bridge (中文)

本地 AI 网关 — 将本地已认证的 AI 工具暴露为 OpenAI 兼容 和 Anthropic 兼容 的 HTTP API。

仅在 127.0.0.1 运行。凭据从现有本地配置文件中读取，不会提取、转发或存储令牌。

快速开始

git clone https://github.com/hotea/LocalAI-Bridge.git
cd LocalAI-Bridge
npm install
npm run dev

或一键安装：

curl -fsSL https://raw.githubusercontent.com/hotea/LocalAI-Bridge/main/install.sh | bash

服务启动于 http://127.0.0.1:6271。

支持的提供商

提供商 方式 认证
Claude Code 子进程（claude CLI） 自动检测 claude --version
GitHub Copilot HTTP API (api.githubcopilot.com) 自动扫描 ~/.config/github-copilot/
Kiro HTTP API (AWS CodeWhisperer) 自动扫描 ~/.aws/sso/cache/
Antigravity HTTP API (Google Cloud AI) 自动扫描 Antigravity IDE 本地数据，或 config.json

前置要求： Node.js >= 20，且至少配置一个本地提供商。

API 端点

方法 路径 说明
GET /health 提供商健康状态
GET /v1/models 列出所有可用模型
POST /v1/chat/completions OpenAI 格式聊天补全（支持流式）
POST /v1/messages Anthropic Messages API（支持流式）

使用示例

健康检查 & 列出模型

# 健康检查
curl http://127.0.0.1:6271/health

# 列出所有可用模型
curl http://127.0.0.1:6271/v1/models

Claude Code CLI

非流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-code:sonnet",
    "messages": [{"role": "user", "content": "用一句话说你好"}]
  }'

流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-code:opus",
    "messages": [{"role": "user", "content": "写一个 Python 函数反转字符串"}],
    "stream": true
  }'

可用模型：claude-code:sonnet, claude-code:opus, claude-code:haiku

GitHub Copilot

非流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot:claude-sonnet-4",
    "messages": [{"role": "user", "content": "解释 JavaScript 中的闭包"}]
  }'

流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot:gpt-4o",
    "messages": [{"role": "user", "content": "用 Go 写一个二分查找算法"}],
    "stream": true
  }'

带系统提示词：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "copilot:claude-sonnet-4",
    "messages": [
      {"role": "system", "content": "你是一个编程助手，回答要简洁。"},
      {"role": "user", "content": "如何反转链表？"}
    ]
  }'

可用模型：copilot:claude-sonnet-4, copilot:claude-3.7-sonnet, copilot:gpt-4o, copilot:gpt-4.1, copilot:o4-mini, copilot:gemini-2.0-flash, copilot:gemini-2.5-pro

Kiro (AWS CodeWhisperer)

非流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kiro:claude-sonnet-4",
    "messages": [{"role": "user", "content": "TCP 和 UDP 有什么区别？"}]
  }'

流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kiro:claude-sonnet-4.5",
    "messages": [{"role": "user", "content": "用 Rust 写一个函数找最长公共子序列"}],
    "stream": true
  }'

多轮对话：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kiro:claude-sonnet-4",
    "messages": [
      {"role": "user", "content": "什么是二叉树？"},
      {"role": "assistant", "content": "二叉树是一种树状数据结构，每个节点最多有两个子节点。"},
      {"role": "user", "content": "如何中序遍历？"}
    ]
  }'

可用模型：kiro:claude-sonnet-4, kiro:claude-sonnet-4.5, kiro:claude-3-7-sonnet, kiro:claude-haiku-4.5

Antigravity (Google Cloud AI)

非流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "antigravity:claude-sonnet-4-5",
    "messages": [{"role": "user", "content": "用简单的语言解释量子计算"}]
  }'

流式：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "antigravity:gemini-3-pro-high",
    "messages": [{"role": "user", "content": "写一个 TypeScript 函数实现防抖"}],
    "stream": true
  }'

带系统提示词：

curl -X POST http://127.0.0.1:6271/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "antigravity:claude-opus-4-6-thinking",
    "messages": [
      {"role": "system", "content": "你是一个资深软件架构师。"},
      {"role": "user", "content": "设计一个可扩展的微服务架构"}
    ]
  }'

可用模型：antigravity:claude-sonnet-4-5, antigravity:claude-sonnet-4-5-thinking, antigravity:claude-opus-4-5-thinking, antigravity:claude-opus-4-6-thinking, antigravity:gemini-3-pro-high, antigravity:gemini-3-pro-low, antigravity:gemini-3-flash

Anthropic Messages API

所有供应商都支持 Anthropic Messages API 格式：

curl -X POST http://127.0.0.1:6271/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: not-needed" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "copilot:claude-sonnet-4",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "你好"}]
  }'

流式：

curl -X POST http://127.0.0.1:6271/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: not-needed" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "antigravity:claude-sonnet-4-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "你好"}],
    "stream": true
  }'

模型命名

使用 provider:model 格式：

供应商 示例模型
claude-code claude-code:sonnet, claude-code:opus, claude-code:haiku
copilot copilot:claude-sonnet-4, copilot:gpt-4o, copilot:gemini-2.5-pro
kiro kiro:claude-sonnet-4, kiro:claude-sonnet-4.5
antigravity antigravity:claude-sonnet-4-5, antigravity:gemini-3-pro-high

内置别名：gpt-4 / gpt-4o → copilot:gpt-4o, gpt-4o-mini / gpt-3.5-turbo → copilot:o4-mini

Antigravity 设置

如果本地已安装 Antigravity IDE，凭据会自动从其数据目录扫描，无需额外配置。

如果自动扫描未找到凭据，可以手动配置：

方式一 — OAuth 登录流程：

  1. 启动服务：npm run dev
  2. curl -X POST http://127.0.0.1:6271/auth/antigravity/login — 返回 Google OAuth 链接
  3. 在浏览器中打开链接并登录
  4. 重定向后刷新令牌自动保存到 config.json
  5. 重启服务

方式二 — 手动配置：

在 config.json 中添加：

{
  "providers": {
    "antigravity": {
      "refreshToken": "1//your-refresh-token-here"
    }
  }
}

配置

复制 config.default.json 为 config.json（已加入 gitignore）进行自定义：

cp config.default.json config.json

环境变量覆盖：

变量 说明
LOCALAI_HOST 绑定地址（默认 127.0.0.1）
LOCALAI_PORT 端口（默认 6271）
LOCALAI_CLAUDE_BINARY claude CLI 路径

许可证

MIT

About

No description or website provided.

Topics

ai-gateway ai-aggregation aiclient2api

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Languages