██████╗ ██████╗  █████╗ ██╗
██╔══██╗██╔══██╗██╔══██╗██║
██████╔╝██║  ██║███████║██║
██╔══██╗██║  ██║██╔══██║██║
██║  ██║██████╔╝██║  ██║██║
╚═╝  ╚═╝╚═════╝ ╚═╝  ╚═╝╚═╝
  

⚡ The Problem

One provider fails (rate limit / crash / timeout)
                ↓
   rdai automatically switches to the next
                ↓
        Your app never goes down

rdai is a self-healing Python SDK that lets you call multiple AI models — Gemini, OpenAI, Claude, DeepSeek, and more — through one unified interface. It auto-discovers your API keys, picks the best provider for your strategy, and silently fails over to a backup the moment something breaks.

🩺 See it in action — rdai doctor

The built-in doctor pings every configured provider live and reports real status, not guesses.

Real provider diagnostics — live authentication, reachability, and latency in one command.

One command tells you exactly which providers are wired up, authenticated, and reachable — before you ship.

📦 Installation

pip install rdai

🚦 Quick Start

rdai init          # interactive wizard → choose providers, strategy, generate .env + rdai.yaml
# add your API keys to the generated .env file
rdai doctor         # verify every provider is alive

GEMINI_API_KEY=your_key_here
GROQ_API_KEY=your_key_here
OPENAI_API_KEY=your_key_here
# add others as selected...

(Your API keys stay strictly local — never hardcoded, never sent anywhere but the provider you're calling.)

💻 Python SDK Usage

No need to learn ten different SDKs — rdai standardizes everything into one call.

from rdai import AI

ai = AI()  # auto-loads your rdai.yaml strategy

response = ai.generate("Write a multi-agent orchestration script in Python.")
print(response)
# Hello! Here's your answer...

If the first model fails, rdai silently falls back to the next one in your chain — no extra code required.

🎯 Overriding Default Models

By default, rdai uses the most stable models for each provider (e.g., gemini-2.5-pro, gpt-4o-mini). However, you have full control to override them by passing a models dictionary:

from rdai import AI

# Override the defaults with your preferred models
ai = AI(models={
    "gemini": "gemini-1.5-flash",
    "groq": "llama3-8b-8192"
})

response = ai.generate("Hello world!")

---

## 🛠️ Bring Your Own Model (BYOM)

Plug a custom or private API straight into the failover chain:

```python
from rdai.providers.base import BaseProvider
from rdai import AI
import requests

class CustomNexusProvider(BaseProvider):
    def __init__(self, api_key):
        super().__init__(api_key, "nexus-v1")

    def generate(self, prompt: str, **kwargs) -> str:
        res = requests.post(
            "https://api.nexus.com/v1",
            headers={"Key": self.api_key},
            json={"text": prompt},
        )
        return res.json()["reply"]

ai = AI(providers=[CustomNexusProvider(api_key="your_custom_key")])
print(ai.generate("Hello Custom Engine!"))

✨ Why rdai?

⚙️ Routing Strategies & Configuration

rdai reads its setup from an optional rdai.yaml in your working directory. Environment variables always take precedence over .env values.

strategy: smart
provider_order:
  - gemini
  - openai
  - groq

• smart — selects the ready provider that best fits the specific request.
• manual — strictly follows the provider_order you define.

Both strategies keep every other ready provider on standby as automatic fallback for transient rate-limit or timeout failures.

🎛️ Supported AI Engines

🛠️ CLI Command Reference

🗺️ Roadmap

• ✅ v1.0.0 — Multi-brain orchestrator core: Unbreakable Auto-Failover, Gemini/OpenAI/Claude/Groq + custom model support, rdai init setup wizard, rdai doctor live diagnostics

• ✅ v1.0.1 — Faster dashboard rendering (loading animation removed), expanded PyPI SEO keywords, corrected GitHub project URLs

• ✅ v1.0.2 — Core architecture unified, missing dependencies resolved, explicit timeouts added for REST providers, and enhanced CLI doctor diagnostics.

• 🟡 v1.1 — Live streaming: brain activity, frontend events, provider timeline

Full version history: CHANGELOG.md

📂 More Examples

Ready-to-run scripts covering failover, BYOM, and each provider live in /examples — clone the repo and run them directly.

git clone https://github.com/ranajitdharpersonal/rdai.git
cd rdai/examples
python basic_usage.py

🤝 Contributing

Issues and PRs are welcome — check open issues or open a new one to discuss a change before submitting a PR.

If rdai saved you from a 3am provider outage, a ⭐ on the repo goes a long way — it's the easiest way to help other developers discover it.