semanticbridge

SemanticBridge SDK

A TypeScript/JavaScript SDK for interacting with the SemanticBridge API. This SDK makes it easy to call the /api/v1/chat/completions endpoint with type safety and a modern developer experience.

Installation

npm install semanticbridge

Usage

Standard (no documents)

import { SemanticBridge } from "semanticbridge";

// The base URL is hardcoded to http://localhost:3030/api
const bridge = new SemanticBridge({
  apiKey: "sk-...",
});

const response = await bridge.ask({
  query: "What are the latest trends in AI?",
  models: ["gpt-4o-mini", "gemini-1.5-flash"],
});

console.log(response.best_response);

With Documents (non-streaming)

const response = await bridge.ask({
  query: "Summarize this document.",
  models: ["gpt-4o-mini", "gemini-1.5-flash"],
  documents: [{ url: "https://example.com/file.pdf" }],
});
console.log(response.best_response);

Streaming with Documents

for await (const token of bridge.askStream({
  query: "Summarize this document.",
  models: ["gpt-4o-mini", "gemini-1.5-flash"],
  documents: [{ url: "https://example.com/file.pdf" }],
  stream: true,
})) {
  console.log("Token:", token.token);
}

Middleware

You can add middleware to intercept or modify requests and responses:

bridge.use(async (req, next) => {
  console.log("Request:", req);
  const res = await next(req);
  console.log("Response:", res);
  return res;
});

Features

• TypeScript-first, fully typed
• Easy to use and extend
• Middleware system for advanced use cases
• Streaming support for real-time tokens
• Document upload support via public URLs
• Fixed API endpoint - Base URL is hardcoded to http://localhost:3030/api

API

• ask(params) — Query the orchestrated LLM endpoint (with or without documents)
• askStream(params, onToken?) — Stream tokens in real time (callback or async iterator)
• use(middleware) — Add a middleware function

License

MIT