python-ai-ragbot

python-ai-ragbot is a modular and framework-agnostic Python package for building intelligent chatbots and voicebots with Retrieval-Augmented Generation (RAG), powered by OpenAI and LangChain.

It provides a simple interface to attach ready-made request handlers into popular frameworks (FastAPI, Flask). You can quickly add both text chat (/chat) and voice (/voice) endpoints into your app. For other frameworks we are aggresively working to release the update

---

Features

• Supports knowledge sources:
  • Local files (.pdf, .docx, .txt, .md)
  • Website scraping (URLs, sitemaps)
• /chat endpoint (text query → answer)
• /voice endpoint (speech-to-text via Whisper, TTS for responses)
• Fully configurable models, embeddings, voices, chunking, logging
• In-memory FAISS vector store via LangChain
• Adapters for:
  • FastAPI
  • Starlette
  • Flask
  • Django
  • Raw WSGI
  • Raw ASGI
• Sync (init_rag_voice_bot) and Async (init_rag_voice_bot_async) APIs

---

Requirements

• Python 3.9+
• An OpenAI API key (OPENAI_API_KEY in .env)

---

Installation

pip install python-ai-ragbot

For local development:

git clone https://github.com/your-org/python-ai-ragbot.git
cd python-ai-ragbot
pip install -e .

---

Quick Start (FastAPI)

# examples/server.py
from fastapi import FastAPI
from python_ai_ragbot import init_rag_voice_bot_async
from python_ai_ragbot.http.adapters import use_in_fastapi
from fastapi.middleware.cors import CORSMiddleware
from dotenv import load_dotenv
import os

load_dotenv()

app = FastAPI()
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

@app.on_event("startup")
async def startup_event():
    bot = await init_rag_voice_bot_async({
        "sources": {"files": ["examples/knowledge.txt"]},
        "openai": {
            "apiKey": os.getenv("OPENAI_API_KEY"),
            "chat": {"model": "gpt-4o"},
            "stt": {"model": "whisper-1"},
            "tts": {"model": "tts-1-hd", "voice": "nova"},
        }
    })
    use_in_fastapi(app, bot["chat_handler"], bot["voice_handler"], prefix="/api/bot")

Run:

uvicorn examples.server:app --reload --port 3001

Endpoints:

• POST /api/bot/chat
• POST /api/bot/voice

---

Usage with Other Frameworks

Flask

from flask import Flask
from python_ai_ragbot import init_rag_voice_bot
from python_ai_ragbot.http.adapters import use_in_flask
import os
from dotenv import load_dotenv

load_dotenv()
app = Flask(__name__)

bot = init_rag_voice_bot({
    "sources": {"files": ["examples/knowledge.txt"]},
    "openai": {"apiKey": os.getenv("OPENAI_API_KEY")},
})
use_in_flask(app, bot["chat_handler"], bot["voice_handler"], prefix="/api/bot")

if __name__ == "__main__":
    app.run(port=3001)

---

Configuration

{
  "sources": {
    "files": ["knowledge.txt", "knowledge.pdf"],
    "urls": ["https://docs.example.com"]
  },
  "rag": {
    "textSplit": {"chunkSize": 1000, "chunkOverlap": 200},
    "topK": 3
  },
  "openai": {
    "apiKey": "...",
    "embeddings": {"model": "text-embedding-3-small"},
    "chat": {"model": "gpt-4o", "temperature": 0.3},
    "stt": {"model": "whisper-1"},
    "tts": {"model": "tts-1-hd", "voice": "nova"}
  },
  "logger": "console"
}

---

Endpoints

/chat

• POST JSON

{"question": "What is in the knowledge base?"}

/voice

• POST raw audio (audio/webm, audio/wav, etc.)

---

Example Project Structure

my-app/
├── examples/
│   ├── server.py
│   ├── knowledge.txt
├── src/
│   └── python_ai_ragbot/
├── .env
└── pyproject.toml

---

Notes

• Use init_rag_voice_bot_async in ASGI frameworks (FastAPI, Starlette).
• Use init_rag_voice_bot in WSGI frameworks (Flask, Django, raw WSGI).
• Vector store is in-memory only; data is reloaded on each startup.

---