Skip to content

messaging.md

Latest commit

 

History

History
145 lines (113 loc) · 3.71 KB

messaging.md

File metadata and controls

145 lines (113 loc) · 3.71 KB

Messaging Channels

Each channel has fully isolated conversation history — no context bleeding between platforms.

Telegram

Set token + whitelist in the Web UI (Settings → Telegram) or via CLI:

/telegram token 123456:ABC...
/telegram add 123456789

Or via .env:

TELEGRAM_BOT_TOKEN=123456:ABC...
TELEGRAM_CHAT_ID=123456789

Telegram Tools:

Tool What it does
send_telegram_message(text) Send text message with Markdown
send_telegram_document(path, caption) Send any file (.docx, .pdf, .py, etc.)
send_telegram_voice(path) Send audio file as voice message
telegram_add_user(chat_id) Add user to allowlist
telegram_list_users() List all allowed users

Supported message types:

  • ✅ Text (with Markdown formatting)
  • ✅ Images (auto-sent when AION generates them)
  • ✅ Voice messages (transcribed offline via Faster Whisper)
  • ✅ Documents & files (any format)
  • ✅ Voice replies (TTS via edge-tts or pyttsx3)

Discord

DISCORD_BOT_TOKEN=...

Responds to DMs and @mentions. Slash command /ask. Enable MESSAGE CONTENT INTENT in the Discord Developer Portal.

Slack

SLACK_BOT_TOKEN=xoxb-...
SLACK_APP_TOKEN=xapp-...

Socket Mode — no public webhook needed. Responds to DMs and @aion mentions.

Amazon Alexa

POST /api/alexa — Alexa Skill endpoint. Configure the HTTPS endpoint in Alexa Developer Console.

Audio

Direction What happens
Voice message → AION Transcribed offline via Faster Whisper (multilingual, no API keys)
AION → voice reply TTS via edge-tts (Microsoft Neural, recommended) or pyttsx3 (offline)
audio_transcribe_any tool Transcribe any audio file format
audio_tts tool Convert text to speech, save as file
Telegram voice OGG OPUS via ffmpeg (auto-converted if needed)
Discord voice MP3/WAV file attachment

STT (Speech-to-Text):

  • Fully offline, no API keys needed
  • Supports OGG, MP3, WAV, M4A, FLAC, WebM
  • Auto-detects language
  • Model size: small (~465 MB), medium, large-v3

TTS (Text-to-Speech):

  • Configure in Web UI (System tab) or config.json
  • Default: edge (best quality, requires internet)
  • Fallback: sapi5 on Windows, pyttsx3 on all platforms
{
  "tts_engine": "edge",
  "tts_voice": "de-DE-KatjaNeural"
}

Browser Automation

AION controls a real Chromium browser via Playwright:

Tool Action
browser_open(url) Load a page
browser_screenshot() Capture full page
browser_click(selector) Click by CSS selector
browser_fill(selector, value) Fill input fields
browser_get_text() Extract page text
browser_evaluate(js) Run JavaScript
browser_find(query) Natural language element search
browser_close() Close the browser

One-time setup:

pip install playwright
python -m playwright install chromium

Configure headless mode: config.json → "browser_headless": true

MCP Client — 1,700+ Integrations

pip install mcp
npm install -g npx

Add servers to mcp_servers.json (no secrets — committable):

{
  "servers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "vault_env": {"GITHUB_PERSONAL_ACCESS_TOKEN": "mcp_github"}
    },
    "postgres": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres"],
      "vault_env": {"POSTGRES_CONNECTION_STRING": "mcp_postgres"}
    }
  }
}

Store secrets in the encrypted vault:

credential_write("mcp_github", "ghp_your_token_here")

Tools are auto-discovered and registered as mcp_{server}_{tool} on startup.