BUDUCCA // OPEN MODEL RELAY CHANNELS LIVE
Notice: BUDUCCA is pre-release software and is not associated with any tokens or financial products. If you hit a problem, join the Telegram or file a GitHub issue.
Messaging-first assistant runtime

Run open models where your messages already live.

BUDUCCA is a lightweight control plane for Telegram, Signal, WhatsApp, and Android assistants. Plug in a modern OpenAI-compatible model endpoint, wire up skills and collectors, and keep state in plain files instead of black-box cloud machinery.

GitHub Telegram Quickstart

Why It Hits Different

  • Messaging-native architecture, not a browser dashboard in disguise.
  • One assistant core across multiple frontends.
  • Skills plus collectors for actions and live context.

Open-Model Ready

The model layer speaks the OpenAI chat-completions protocol, so BUDUCCA fits cleanly behind local or remote open-model serving stacks.

Plain-File State

Traces, collector outputs, unanswered messages, and learned context stay in a readable workspace. That keeps debugging, backup, diffing, and manual inspection straightforward when you need to see exactly what the assistant has stored. Audit trail included.

Minimal Surface Area

Small Python modules, direct config files, and no required hosted backend. The stack is easier to reason about because most changes happen close to the behavior they affect, without hidden orchestration or extra service glue. Fewer layers. Less drift. Faster edits.

Quickstart

Clone, point `llm.json` at your model endpoint, enable one frontend, then boot the relay.

cp -R config.example config
$EDITOR config/llm.json
$EDITOR config/telegram.json
python3 run_bot.py --config config

For open-model serving, start with LM Studio or Ollama's OpenAI-compatible API.

Getting Started Complete setup, local commands, collector boot, and in-chat commands. AI Stack How BUDUCCA fits together and how it compares with hosted assistants, agent frameworks, RAG bots, and automation platforms. Agent Modernization Current backend profiles, recent agent patterns worth adopting, explicit exclusions, and the efficiency-first delivery plan. Frontends Telegram, Signal, WhatsApp, Android, and the built-in slash command surface. Developer Guide Architecture, plugin layout, config loading, and extension points.