The trickle -- lightweight LLM agent infrastructure by ellmos-ai.
Lightweight LLM agent infrastructure: Memory, Connectors, Automation.
Extracted from BACH -- a comprehensive agent system with 73 handlers, 322 tools, and 24 protocols. Rinnsal takes only the infrastructure layer and leaves agent/skill/tool logic to the LLM providers.
- Memory -- Cross-agent shared memory with SQLite (facts, working memory, lessons learned, sessions)
- Tasks -- Simple task management with priorities, status tracking, and agent assignment
- Connectors -- Channel abstraction for Telegram, Discord, Home Assistant
- Automation -- LLM agent chain orchestration ("Marble-Run": sequential agent chains with loops, handoff, shutdown conditions)
- Ollama -- Local LLM runner for Ollama REST API (qwen3, mistral, etc.)
- Zero dependencies -- Pure Python stdlib, no external packages required
- Python 3.10+
pip install -e .from rinnsal.memory import api
api.init(agent_id="my-agent")
api.fact("system", "os", "Windows 11")
api.note("Aktueller Task: Feature implementieren")
api.lesson("UTF-8 Bug", "cp1252 Encoding", "PYTHONIOENCODING=utf-8", severity="high")
print(api.context()) # Kompakter Kontext fuer LLM-Prompts
print(api.status()) # Statistikenfrom rinnsal.tasks import api as tasks
tasks.init(agent_id="my-agent")
tasks.add("Implement feature X", priority="high", description="Details here")
tasks.add("Fix encoding bug", priority="critical")
for t in tasks.list():
print(f"[{t['id']}] {t['title']} ({t['status']})")
tasks.activate(1) # Set to 'active'
tasks.done(1) # Mark as done
print(tasks.next_task()) # Next open task by priorityfrom rinnsal.auto.ollama_runner import OllamaRunner
ollama = OllamaRunner(model="qwen3:4b", base_url="http://localhost:11434")
if ollama.health():
result = ollama.run("Explain this code")
print(result["output"])
# Chat with message history
result = ollama.chat([
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello!"}
])
print(result["output"])
print(ollama.available_models()) # List installed modelsfrom rinnsal.connectors import load_connector
# Token via ENV: export RINNSAL_TELEGRAM_TOKEN=123:ABC...
tg = load_connector("telegram")
tg.connect()
tg.send_message("chat_id", "Hallo von Rinnsal!")from rinnsal.auto.runner import ClaudeRunner
runner = ClaudeRunner(model="claude-sonnet-4-6")
result = runner.run("Analysiere den Code in src/")
print(result["output"])rinnsal status # Gesamtstatus
rinnsal version # Version
# Memory
rinnsal memory status # Memory-Statistiken
rinnsal memory fact system os "Win 11" # Fakt speichern
rinnsal memory facts --json # Alle Fakten (JSON)
rinnsal memory note "Meine Notiz" # Notiz speichern
rinnsal memory context # LLM-Kontext generieren
# Tasks
rinnsal task add "My task" -p high # Create task (critical/high/medium/low)
rinnsal task add "Bug fix" -d "Details" # With description
rinnsal task list # Open/active tasks
rinnsal task list --all # Including done/cancelled
rinnsal task list --json # JSON output
rinnsal task show 1 # Task details
rinnsal task done 1 # Mark as done
rinnsal task activate 1 # Set to active
rinnsal task cancel 1 # Cancel task
rinnsal task reopen 1 # Reopen done/cancelled
rinnsal task delete 1 # Delete permanently
rinnsal task count # Count by status
# Chains (Automation)
rinnsal chain list # Ketten auflisten
rinnsal chain start mein-projekt # Kette starten
rinnsal chain status # Status aller Ketten
rinnsal chain stop mein-projekt # Kette stoppen
rinnsal chain log mein-projekt # Log anzeigen
rinnsal chain reset mein-projekt # Zuruecksetzen
rinnsal chain create # Interaktiv erstellen
# Connectors
rinnsal connect list # Verfuegbare Connectors
rinnsal connect test telegram # Verbindungstest
rinnsal connect send telegram ID "Text" # Nachricht senden
# Pipe (Einzelaufruf)
rinnsal pipe "Erklaere diesen Code" # Einzelner LLM-AufrufRinnsal sucht die Config in folgender Reihenfolge:
./rinnsal.json(aktuelles Verzeichnis)~/.rinnsal/config.json(User-Home)
Secrets werden ausschliesslich via ENV-Variablen gesetzt:
export RINNSAL_TELEGRAM_TOKEN="123456:ABC-DEF..."
export RINNSAL_DISCORD_TOKEN="..."
export RINNSAL_HA_TOKEN="..."Beispiel-Config: siehe config/rinnsal.example.json
rinnsal/
├── memory/ # SQLite-based cross-agent memory (from USMC)
├── tasks/ # Task management with SQLite (priorities, status)
├── connectors/ # Messaging channel abstraction (from BACH)
├── auto/ # LLM agent chain orchestration + OllamaRunner
└── shared/ # Config loader, event bus
Memory <-> Auto: Chains read/write context from Memory (opt-in)
Connectors <-> Auto: Telegram notifications via Connector
Event Bus: Decoupling layer between components
How does Rinnsal compare to OpenClaw, the popular open-source AI assistant (274K+ stars)?
| Rinnsal | OpenClaw | |
|---|---|---|
| Focus | Minimal infrastructure layer -- Memory, Connectors, Automation | Full AI assistant -- 20+ messengers, native apps, voice, skill marketplace |
| Philosophy | Provide building blocks, let the LLM handle the rest | All-in-one ecosystem with community-driven extensions |
| Memory | Structured SQLite (facts, lessons, working memory, sessions) | Session-based with /compact, workspace files |
| Connectors | Telegram, Discord, Home Assistant (same abstraction pattern, growing) | 20+ platforms (WhatsApp, Slack, Signal, Teams, Matrix...) |
| Dependencies | Zero -- pure Python stdlib | Node.js 22+, numerous npm packages |
| License | MIT | MIT |
In short: Rinnsal and OpenClaw share the same connector gateway pattern, but take opposite approaches to complexity. Rinnsal stays minimal with zero dependencies and structured memory. OpenClaw goes all-in with native apps, voice, and a massive community. Different starting points, potentially converging on connectors.
MIT -- Lukas Geiger