WebAI-to-API

WebAI-to-API is a browser-native AI runtime that exposes browser-based AI services through OpenAI-compatible APIs.

WebAI-to-API combines browser-native automation with WebAPI-based provider integrations to expose AI services through a flexible OpenAI-compatible API layer.

Features

OpenAI-compatible /v1/chat/completions API
Provider-based architecture with unified routing
Streaming response support (SSE)
Conversation continuation support
Health, readiness, and runtime diagnostics endpoints
Docker deployment support
Authentication management and browser login workflows

Available Providers

Gemini

Provides access to Google Gemini models through either the WebAPI backend or a browser-native Playwright runtime.

Atlas Cloud

Provides access to cloud-hosted AI models through a native API integration powered by Atlas Cloud.

Quick Start

1. Install Dependencies

bash
poetry install
poetry run playwright install chromium

2. Configuration

bash
cp config.conf.example config.conf

For detailed settings (including logging verbosity and access log configurations), see Configuration Guide.

3. Authenticate

bash
python verify_login.py

4. Start the Server

bash
poetry run python src/run.py

[!TIP]
New users can run make setup and make doctor for automated setup and diagnostics. See Convenience Shortcuts below.

Optional: Convenience Shortcuts

WebAI-to-API includes a bootstrap utility and a Makefile for common setup tasks.

Command	Description
`make setup`	One-step install, directory creation, and config setup.
`make doctor`	Run environment and dependency diagnostics.

Alternative (no Make): python scripts/bootstrap.py and python scripts/doctor.py

Authentication

Gemini requires an authenticated Google session.

Method	Recommended For
Browser Login (`verify_login.py`)	Recommended. Playwright backend, Docker, and long-term usage.
Manual Cookies	Quick testing and WebAPI-only usage.

Run the interactive login helper:
```
bash
python verify_login.py
```
Complete the sign-in process in the browser window.
This creates runtime/auth/gemini.json, used by both backends.

2. Manual Cookies

Sign in to Gemini.
Copy __Secure-1PSID and __Secure-1PSIDTS from your browser cookies.
Paste them into the [Gemini] section of config.conf.

Send Your First Request

bash
curl -X POST http://localhost:6969/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'

Dashboard

Open the dashboard at http://localhost:6969/ui.

The dashboard provides a visual interface for runtime status, authentication management, API discovery, and interactive testing.

Available Endpoints

Endpoint	Purpose
`/v1/chat/completions`	OpenAI-compatible chat completions
`/v1/temporary/chat/completions`	Gemini WebAPI temporary chat
`/v1/models`	List supported models
`/v1/auth/status`	Check authentication status
`/v1/auth/login`	Start authentication flow
`/v1/conversations`	Manage Gemini WebAPI conversation snapshots
`/translate`	Legacy translation compatibility
`/ui/*`	Dashboard and playground

Temporary Gemini Chat

/v1/temporary/chat/completions is Gemini WebAPI-only and uses Gemini temporary requests (temporary=True).
Requests are not stored in Gemini history and do not create SQLite conversation snapshots.
It supports streaming, non-streaming, multimodal file inputs, and artifact outputs. See docs/api.md for details.

Legacy Translate Endpoint

The /translate endpoint is maintained for compatibility with the Translate It! browser extension.

It uses Gemini temporary requests and a shared in-memory session. See docs/api.md for additional details.

File Support

File input is supported through OpenAI-style content parts on /v1/chat/completions when routed to the Gemini WebAPI backend.
The currently verified file formats are documented in docs/api.md.

[!NOTE]
File parts are supported only by the Gemini WebAPI backend in the MVP. Gemini Playwright and Atlas reject file parts with a clear capability error.
For Gemini WebAPI, text content parts are concatenated into one prompt and file parts are passed as attachments. Exact text/file interleaving order is not preserved.
The built-in /ui/playground page uses the same contract for file attachments.

Supported Models

Available models may vary depending on the configured provider and backend.

Use the /v1/models endpoint to retrieve the current list of supported models.

Model Routing

WebAI-to-API uses model prefixes to route requests to specific backends.

Model	Backend
`gemini-3-flash`	Gemini (default configured backend)
`playwright/gemini-3-flash`	Gemini Playwright
`atlas/<model-id>`	Atlas Cloud via OpenAI-compatible API, with 50 validated chat models exposed in `/v1/models`

[!TIP]
Model prefixes force backend selection and override the default Gemini backend configured in config.conf. Use playwright/... model prefixes to force the Playwright backend explicitly.