Remote control for astraeus — send orders from anywhere, get agent responses back.
Send a text message on Telegram → astraeus agent processes it → response sent back to Telegram.
Persistent sessions mean the agent remembers your conversation. Zero cost when idle.
- DeepWiki GitHub reference capture for repository URLs
- Voice message transcription via
faster-whisper - Image OCR and PDF extraction into
.sisyphus/references/ - Multi-project routing using
#projecttags and keyword fallback - Priority and scheduling markers (
🔴,🔵,⏰2h) - Optional digest push scheduler with SQLite persistence
# 1. Load into your astraeus project
astraeus load galaxy-protocol
# 2. Create bot via @BotFather on Telegram
# 3. Configure
cp galaxy-protocol/tools/config.json.example .galaxy/config.json
# Edit with your telegram_token and authorized_users
# 4. Install deps
pip install -r galaxy-protocol/tools/requirements.txt
# Optional system deps for OCR/media
# Ubuntu/Debian: sudo apt install ffmpeg tesseract-ocr
# macOS: brew install ffmpeg tesseract
# 5. Run
python3 galaxy-protocol/tools/bot.py &
python3 galaxy-protocol/tools/hermes.py --interval 5 &
# Optional: validate enabled feature dependencies
python3 galaxy-protocol/tools/feature_preflight.pyPhone → Telegram → bot.py → order file → Hermes → opencode run → response → Telegram
| Component | Description |
|---|---|
tools/bot.py |
Telegram bot (any text = order) |
tools/hermes.py |
Order dispatcher (5s polling, 3min timeout, persistent sessions) |
tools/galaxy_mcp.py |
MCP server for programmatic access |
services/ |
systemd units for production deployment |
- astraeus >= 0.3.0
- Python 3.10+
python-telegram-bot(see requirements.txt)
Feature flags live in .galaxy/config.json under features and are disabled by default.
GALAXY_DEEPWIKI_ENABLEDGALAXY_VOICE_ENABLEDGALAXY_IMAGE_PDF_ENABLEDGALAXY_MULTI_PROJECT_ENABLEDGALAXY_PRIORITY_SCHEDULING_ENABLEDGALAXY_DIGEST_PUSH_ENABLED
Same as astraeus.
Transform GitHub repository references into actionable intelligence with architectural insights and adoption paths.
- Deep Analysis: Extracts architecture, design patterns, and core abstractions from GitHub repos via DeepWiki MCP
- Diagnostic Intelligence: Generates concrete adoption paths with complexity indicators (🟢🟡🔴)
- Scheduled Push: Daily 9 AM KST digest to Telegram (optional)
- Backward Compatible: Feature-flagged, falls back to shallow capture if disabled
Enable in .galaxy/config.json:
{
"features": {
"GALAXY_DEEPWIKI_ENABLED": true,
"GALAXY_DIGEST_PUSH_ENABLED": true
}
}# Capture GitHub repo with deep analysis
/feed https://github.com/vercel/ai
# Generate diagnostic digest
/digest --last 7| Indicator | Meaning | Criteria |
|---|---|---|
| 🟢 Easy | Quick win | ≤1 file, <2 hours, no deps |
| 🟡 Moderate | Planned effort | 2-5 files, 2-8 hours, minor deps |
| 🔴 Complex | Strategic initiative | >5 files, >8 hours, architectural |