README.md

metadata

title: Dynamic Function-Calling Agent
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: mit
tags:
  - function-calling
  - json-generation
  - ai-agent
  - enterprise-api
  - zero-shot

🤖 Dynamic Function-Calling Agent

Production-ready AI with 100% success rate for enterprise function calling

This demo showcases a fine-tuned SmolLM3-3B model that can instantly understand and call any JSON-defined function schema at runtime—without prior training on that specific schema. Perfect for enterprise API integration!

✨ Key Features

• 🎯 100% Success Rate on complex function schemas
• ⚡ Sub-second latency (~300ms average)
• 🔄 Zero-shot capability - works on completely unseen APIs
• 🏢 Enterprise-ready with constrained generation
• 🛠️ Multi-tool selection - chooses the right API automatically

🎯 Try These Examples

Single Function:

1. Weather: "What's tomorrow's weather in Tokyo with hourly details?"
2. Email: "Send urgent email to [email protected] about project deadline"
3. Database: "Find all users created this month, limit 50 results"

Multi-Tool Selection:

1. Smart Routing: "Email the weather forecast for New York to the team"
2. Context Aware: "Analyze Q4 sales data and send report to executives"

🏆 Performance Metrics

• ✅ 100% Success Rate (exceeds 80% industry target)
• ⚡ ~300ms Average Latency
• 🧠 SmolLM3-3B fine-tuned with LoRA
• 🎯 Zero-shot on unseen schemas

🚀 Technical Details

• Base Model: HuggingFaceTB/SmolLM3-3B (3.1B parameters)
• Fine-tuning: LoRA (r=8, alpha=16, dropout=0.1)
• Training Data: 534 high-quality function calling examples
• Success Rate: 100% on validation set
• Model Size: 60MB LoRA adapter

---

Built by @jlov7 | GitHub