🧠 Agent Skill SDK

Build modular, intelligent agents with event triggers, retry logic, memory, LLM planning (Amazon Bedrock), and a live dashboard.

---

🚀 Features

• ✅ Skill-based programming with @skill decorators
• 🔁 Retry, backoff, and fallback on skill failure
• 🧠 LLM reasoning via Amazon Bedrock
• 🧩 Triggers: Timer, MQTT, HTTP, Cron, File Watcher, CoAP
• 📜 Memory tracing + export (CSV/JSON)
• ⚙️ CLI and 🧪 FastAPI-based dashboard
• 📚 Full developer documentation with MkDocs

---

📦 Install

pip install agent-skill-sdk

---

🧪 Example: Define a Skill

from agent_skill_sdk import skill

@skill(trigger="on_temp_high", retries=2, backoff="exponential")
def turn_on_fan(context):
    print("Turning on fan:", context.get("location"))

---

⚡ Trigger from CLI

agent register examples.fan_control.turn_on_fan
agent trigger on_temp_high --data '{"location": "lab"}'

---

🧠 LLM Planning (Amazon Bedrock)

@skill(trigger="on_goal")
def bedrock_planner(context):
    response = run_reasoning(context.get("goal"))
    action = response.get("action")
    context.agent.trigger(action)

Supports:

• Prompt templates
• Memory recall
• Fallback actions

---

🌐 Dashboard

uvicorn dashboard.app:app --reload

Visit http://localhost:8000 to:

• View registered skills
• Trigger events manually
• Inspect LLM trace logs

---

🔌 Trigger Plugins

Trigger	File	Event
Timer	timer.py	on_timer
MQTT	mqtt.py	on_mqtt_message
HTTP/Webhook	http.py	on_http
Cron	cron.py	on_cron
File Watcher	file_watcher.py	on_file_change
CoAP (IoT)	coap.py	on_coap

---

📚 Documentation

Visit your deployed docs here:
➡️ https://subhashtalluri.github.io/agent-skill-sdk/

Or build locally:

mkdocs serve

---

📤 Deployment

Use Docker or deploy with Uvicorn:

uvicorn dashboard.app:app --host 0.0.0.0 --port 8000

---

🧠 Trace Export

agent.export_trace("json")
agent.export_trace("csv")

---

🧩 Contributions Welcome

Build a trigger plugin, skill library, or submit a PR!

---

📜 License

MIT – do anything, just give credit ✨