Go Telegram bot

Receives incoming Telegram messages via polling or webhook, forwards them to OpenAI's ChatGPT API, and streams responses back using Server-Sent Events (SSE). The bot maintains a message processing loop that captures user input, sends it to ChatGPT's streaming endpoint, and progressively updates the Telegram message as tokens arrive, reducing perceived latency compared to waiting for full response completion.

Launches a headless browser window when no stored session token exists, guides the user through OpenAI's login flow, automatically extracts the session token from browser cookies, and persists it to the local config file for future use. This eliminates manual token extraction and handles session refresh transparently, supporting both interactive setup and programmatic authentication.

Manages bot configuration through a hybrid approach: environment variables (.env file) for runtime settings like TELEGRAM_TOKEN and EDIT_WAIT_SECONDS, combined with persistent JSON storage for stateful data like OpenAI session tokens. Configuration is loaded on startup, with environment variables taking precedence, and persistent state is written back to platform-specific config directories after authentication or updates.

Buffers streaming ChatGPT tokens and updates the Telegram message at configurable intervals (default ~1 second via EDIT_WAIT_SECONDS) rather than on every token, respecting Telegram's rate limits (~1 edit per second per message). This prevents API throttling errors and reduces network overhead while maintaining perceived real-time streaming by batching multiple tokens into single edit operations.

Optionally restricts bot access to a single Telegram user by checking the incoming message sender's ID against a configured TELEGRAM_ID value. When set, only messages from that user ID are processed; all others are silently ignored. This provides a simple access control mechanism without requiring a full authentication system, suitable for personal bot deployments.

Provides pre-compiled binaries for macOS (Intel/ARM), Linux (x86/ARM), and Windows, eliminating the need for users to compile from source. Additionally offers a Docker image (ghcr.io/m1guelpf/chatgpt-telegram) that bundles the binary with runtime dependencies, allowing deployment via container orchestration with volume mounts for persistent config and environment variable injection for secrets.

Processes each incoming Telegram message independently without maintaining conversation history or context between messages. Each message is sent to ChatGPT as a standalone request, and responses are isolated to that single message. This stateless design simplifies deployment and avoids memory leaks from unbounded conversation history, but requires users to provide full context in each message if they want multi-turn conversations.

Communicates with OpenAI's ChatGPT API using session-based authentication (session tokens extracted from browser cookies) rather than API keys. Sends user messages to OpenAI's streaming endpoint, receives Server-Sent Events (SSE) with token-by-token responses, and handles streaming response parsing. This approach mirrors the web interface's authentication mechanism, avoiding API key management and supporting the same session lifecycle as the browser.