kalibr 1.0.25

Multi-Model MCP SDK — deploy to GPT, Claude, Gemini, Copilot from one codebase.

Kalibr SDK

Multi-Model AI Integration Framework

Write once. Deploy anywhere. Connect to any AI model.

Kalibr turns Python functions into APIs that work seamlessly with GPT, Claude, Gemini, and Copilot — automatically generating model-specific schemas and endpoints.

---

🚀 Quick Start (2 minutes)

1. Install

pip install kalibr

2. Get Examples

kalibr-connect examples

This copies example files to ./kalibr_examples/ in your current directory.

3. Run Demo

kalibr-connect serve kalibr_examples/basic_kalibr_example.py

4. See All Schemas

Kalibr now auto-detects your environment and generates the correct base URLs.

Environment	Example Base URL
Local Dev	http://localhost:8000
Fly.io	https://<app-name>.fly.dev
Custom Host	Use KALIBR_BASE_URL env var

Then open:

<your-base-url>/gpt-actions.json    # ChatGPT
<your-base-url>/mcp.json            # Claude
<your-base-url>/schemas/gemini      # Gemini
<your-base-url>/schemas/copilot     # Copilot

---

🧠 What Kalibr Does

Kalibr turns your Python functions into production-ready multi-model APIs.

from kalibr import Kalibr

app = Kalibr(title="Inventory API")

@app.action("get_inventory", "Fetch inventory data")
def get_inventory(product_id: str):
    return {"product_id": product_id, "stock": 42}

Result:
ChatGPT, Claude, Gemini, and Copilot can all call get_inventory() using their native protocols — no schema work required.

---

💪 Two Modes

Function-Level (Simple)

Ideal for one-off APIs or scripts.

from kalibr import Kalibr

app = Kalibr(title="My API")

@app.action("calculate_price", "Calculate price total")
def calculate_price(product_id: str, quantity: int):
    return {"total": quantity * 19.99}

App-Level (Advanced)

Use KalibrApp for complete control — file uploads, sessions, streaming, and workflows.

from kalibr import KalibrApp
from kalibr.types import FileUpload, Session

app = KalibrApp(title="Advanced API")

@app.file_handler("analyze_doc", [".pdf", ".docx"])
async def analyze_doc(file: FileUpload):
    return {"filename": file.filename, "analysis": "..."}

@app.session_action("save_data", "Save session data")
async def save_data(session: Session, data: dict):
    session.set("my_data", data)
    return {"saved": True}

---

📚 Examples Included

After running kalibr-connect examples, you’ll get:

• basic_kalibr_example.py – simple function-level demo
• enhanced_kalibr_example.py – full app with sessions, uploads, and streaming

---

🤖 AI Platform Integration

ChatGPT (GPT Actions)

1. Copy schema URL:
   https://<your-domain>/gpt-actions.json
2. In GPT Builder → Actions → Import from URL
3. Done — ChatGPT can call your endpoints.

Claude (MCP)

Add to Claude Desktop config:

{
  "mcp": {
    "servers": {
      "my-api": {
        "url": "https://<your-domain>/mcp.json"
      }
    }
  }
}

Gemini / Copilot

Use:

https://<your-domain>/schemas/gemini
https://<your-domain>/schemas/copilot

---

🎯 Common Use Cases

• Customer Service APIs — let AI handle orders or refunds
• Data Analysis — query your analytics through AI
• Document Processing — parse or summarize uploaded docs
• Business Automation — trigger internal workflows
• Internal Tools — expose secure internal logic to assistants

---

🔧 CLI Reference

kalibr-connect examples          # Copy examples
kalibr-connect serve my_app.py   # Run locally
kalibr-connect version           # Show version
kalibr-connect --help            # Full CLI

---

⚡ Key Features

✅ Multi-Model Support — GPT, Claude, Gemini, Copilot
✅ Automatic Schema Generation
✅ Environment-Aware Base URLs (v1.0.21+)
✅ File Uploads
✅ Session Management
✅ Streaming Responses
✅ Workflow Support
✅ Type-Safe API Generation
✅ Async / Await Ready

---

🔥 Why Kalibr?

Without Kalibr:

• Learn 4 model specs
• Maintain 4 codebases
• Duplicate effort

With Kalibr:

• One Python function
• Four schemas generated automatically
• Deploy anywhere

---

🆕 Version 1.0.21+

• Automatic Base-URL Detection
  • Works with KALIBR_BASE_URL or FLY_APP_NAME
  • Fixes all localhost references in deployed schemas
• Ready for MCP ecosystem production use
• Drop-in backwards compatibility

---

🧩 License

MIT License — see LICENSE file for details.

---

Kalibr SDK — the unified layer between AI models and the real world.
Write once. Deploy anywhere. Integrate everything.