PAI RAG Chatbot Builder

Knowledge Documents

Upload PDF files to train your chatbot. The system will parse them and split them into searchable chunks.

📤

Drag & Drop PDF Here

or browse files

PDF files only (max 15MB)

Uploaded Files

📁

No documents uploaded yet. Upload a PDF above to get started.

LLM Endpoint & RAG Settings

Configure the AI model backend that generates customer support responses.

LLM Provider

Model Name

API Base URL Must be the OpenAI-compatible endpoint route. For local Ollama, use: http://localhost:11434/v1

API Key / Token

Widget Chat Title

Allowed Domains (CORS Lock) Comma separated list of domains allowed to load the widget. Use * to allow all.

System Prompt & RAG Context Injection Make sure to include {context} in the prompt where search chunks from PDFs should be injected.

Embed your Chatbot

Copy and paste this script tag at the bottom of your HTML pages (just before the closing </body> tag) to display the RAG chatbot.

HTML Integration Code

This snippet loads the widget script and applies it automatically using your configured settings.

<script src="http://localhost:8000/widget.js" defer></script>

Customization & Options

The embed script handles layout and responsiveness automatically. By default, it places a floating chat button at the bottom right. You can place the widget on any site, including Webflow, WordPress, Shopify, or plain HTML sites.

🔒

CORS Domain Lock

Secured by default. The chat will refuse to respond if loaded on unapproved domains.

⚡

Auto-URL Integration

The widget communicates back to this exact server host automatically without hardcoding ports.

Chatbot Playground

Test your chatbot here. Messages will run through the FTS5 search indexing and query your configured LLM endpoint.

Hello! I am ready to test your PDF knowledge base. Ask me any question!

Last Retrieved RAG Context

Shows the top sentences retrieved from SQLite for the last message.

No context loaded yet. Send a message to inspect RAG retrieval.

Analytics & Self-Learning

Track user questions, collect helpfulness ratings, and add manual FAQ entries to fill knowledge gaps.

Conversation History Logs

Loading logs...

🧠 Fill Knowledge Gaps

Add a custom fact, correction, or Q&A directly. It will be indexed immediately in the search engine.