AI Models Bridge

A TypeScript library that provides a unified interface for interacting with various AI models in browser extensions and web applications.

🚧 Work in Progress

This project is currently under active development. Features and APIs may change without notice. This README, documentation, and installation instructions are also a work in progress.

📋 Overview

AI Models Bridge is a library that allows developers to easily integrate multiple AI models into their browser extensions or web applications. It provides a consistent API for interacting with different AI services, handling the complexity of different authentication methods, request formats, and response parsing.

🧩 Project Structure

The project consists of two main parts:

1. Library: The Typescript library itself, that provides a unified interface for interacting with various AI models
2. Demo Extension: A browser extension that demonstrates how the library can be used

🤖 Supported AI Models and Implementation Status

📊 Implementation Readiness Levels

The following classification system is used to indicate the implementation status of each AI model and identify which features are available for each one.

Level	Name	Description
L1	Basic Text Support	Basic implementation with text-only conversation support. Authentication* and core messaging functionality work, but may have limitations.
L2	Thread Handling Support	Support for conversation history, thread and message management, and conversation state persistence.
L3	Image Support*	Support for images in conversations. Includes proper handling of image uploads and saving and returning data for display/usage.
L4	Model Specific Features	Support for model-specific features like suggested responses, different conversation modes (chat, reasoning, etc.), and streaming responses*.
L5	Improved Error Handling	Slightly improved error handling for AI models.
L6	Multimedia Support	Support for images_and_ other media types in conversations*. Includes mostly proper handling of media uploads and saving and returning data for display/usage.
L7	Enhanced Text Formatting and Support	Improved text handling with support for different model output styles & formats (widgets, code, markdown, tables, etc.). Includes proper rendering and parsing of formatted outputs.
L8	Advanced Features	Advanced features like dynamic model switching*, comprehensive implementation of model-specific capabilities, etc.
L9	Beta Testing/Usage Ready	Implementation ready for testing. Includes and requires comprehensive error handling, rate limiting and quota handling support, proper documentation and testing framework, security measures, caching, extensive unit and integration tests (possibly), detailed performance monitoring and logging for testing, cross-browser compatibility verification, preliminary security audits, final optimizations and extensive real-world testing across different scenarios.
L10	Production Ready (Stable)	Fully tested implementation with comprehensive error handling, rate limiting support, proper documentation, robust security measures, and optimized performance. Ready for production use.

*if supported by the AI model

🎯 Current Model Implementation Status

Model	Status	Details	Implemented Models
Google Gemini Web (Bard)	L6	Text and image support (other file types not supported by model) with conversation history and thread management. Basic error handling implemented. Server operation utility methods, like deleting, sharing, renaming conversations, etc. Response Streaming not supported by model.	2.0 Flash
Copilot Web (Bing Chat)	L5	Supports text and image inputs. Implements conversation modes (chat, reasoning). Has automatic thread title generation, suggested replies, and response streaming. Thread management and conversation history functional. Still requires improved error handling and reconnection logic.	Default (based on GPT-4, GPT-4o and GPT-o3-mini)
Claude Web (Claude)	L5	Supports text and image inputs. Supports automatic and manual thread title generation, response streaming, and Claude specific conversation/response styles. Has a lot of other server operation utility methods, like deleting, sharing, renaming conversations, etc. Thread management and conversation history functional. Still requires somewhat improved error handling and reconnection logic, and support of other models.	3.7 Sonnet
Perplexity Web (Perplexity)	L6	Supports text and image, and other attachment/file inputs. Supports automatic thread title generation, response streaming, and Perplexity specific search focus and sources. Also supports model switching. Server operation utility methods, like deleting, sharing, renaming conversations, etc. Thread management and conversation history functional. Has improved error handling, but still requires better reconnection logic, etc.	Perplexity Sonar, Perplexity Pro Auto, Perplexity Sonar Pro, GPT-4.1, Claude 3.7 Sonnet, Gemini 2.5 Pro, Grok 3 Beta, Perplexity R1 1776, GPT-o4-min