500nits

Botify

A self-hosted bridge between conversational interfaces and repository-based AI agents.

It started in October 2025 with a simple objective: allow a developer to interact with Codex remotely through Telegram while keeping the agent connected to the files, instructions, tools, and working directory of a real project.

As the project evolved, Botify became a reusable agent runtime rather than a basic chatbot integration. Each deployment can be attached to a specific repository or business environment, allowing the agent to work with local project context instead of operating inside an isolated conversation.

Current capabilities include:

• Telegram access through private chats and groups
• Independent, persistent conversations for each authorized chat
• Document, image, audio, and voice-note ingestion
• Local Whisper-based voice transcription
• PostgreSQL-backed shared and searchable memory
• Project-scoped file access and repository awareness
• Operational diagnostics, authentication checks, model controls, and deployment commands
• Automatic startup and recovery on macOS and Debian hosts
• Extensible transports, services, and AI-provider boundaries
• Starter configurations for adapting the system to domain-specific workflows

Botify's architecture is also evolving from an initial monolithic bridge into clearly separated components for message ingestion, command routing, session lifecycle, prompt execution, AI-provider communication, memory, attachments, transcription, and transport delivery.

Timeline

1. Oct 2025 — Project start

   Initial commit. A TypeScript bridge that lets Telegram messages reach Codex through MCP.

2. Nov–Dec 2025 — Usability & collaboration

   Improved status and lifecycle tools, startup scripts, Telegram group support, attachment saving, auth health checks, helper server, and richer formatting.

3. Jan 2026 — Persistent context

   Added PostgreSQL-backed shared memory, per-chat persistent conversation IDs, sender metadata in prompts, diagnostics, and text-to-speech utilities.

4. Feb–Mar 2026 — Multimodal features

   Added audio and voice-note capture, Whisper-based transcription, model controls, attachment-path awareness, and on-demand shared memory queries.

5. Apr 2026 — Major architecture refactor

   Modularized the Telegram bridge and separated concerns into providers, services, command routing, sessions, transcription, memory, and helper-server components.

6. May–Jun 2026 — Expansion & stabilization

   Explored a multi-provider direction and continued hardening the runtime with compatibility, path, and startup refinements.