SmartSupport AI

Open-Source AI Customer Support  ·  RAG  ·  FAISS  ·  SmolLM2  ·  One Script Tag

Add your documents → Build the index → Embed on any website in minutes.

Quick Start  ·  Live Demo  ·  Documentation  ·  Issues

---

Summary

SmartSupport AI is an open-source AI customer support tool that any business can clone, train on their own documents, and embed on their website with no AI experience required and no monthly fees.

It uses Retrieval-Augmented Generation (RAG): your documents are split into chunks, converted to vector embeddings, and stored in a FAISS index. When a customer asks a question, the most relevant chunks are retrieved and passed to a local language model to generate a precise, grounded answer.

Unlike SaaS chatbot tools: you own the data, the model runs on your server, there are no per-query costs, and the entire stack is free to self-host.

Features

	Feature	Description
	RAG Pipeline	FAISS vector search + sentence-transformers for accurate, grounded answers
	Folder-based Setup	Drop files in documents/, run one command - no coding required
	No Database	All knowledge stored as files on disk - zero infrastructure to set up
	One-line Embed	A single <script> tag installs the full chat widget on any website
	Full Customization	Bot name, brand colors, welcome message, and position - all via .env
	Multi-format Support	Plain text, Markdown, and PDF documents supported
	Streaming Responses	Token-by-token streaming via Server-Sent Events
	REST API	Clean FastAPI backend with health check and chat endpoints

Project Structure

smartsupport-ai/
│
├── documents/                  ← PUT YOUR BUSINESS FILES HERE
│   ├── faq.txt                   Your questions and answers
│   └── products.txt              Your product catalog
│
├── backend/
│   ├── main.py                   FastAPI server - /api/chat, /widget.js, /health
│   └── rag_engine.py             Chunking, embedding, FAISS search, generation
│
├── widget/
│   └── widget.js                 Embeddable chat widget - pure JS, zero dependencies
│
├── scripts/
│   └── build_index.py            Run this once after adding your documents
│
├── index/                      ← AUTO-CREATED - do not edit manually
│   ├── faiss.index               Vector index (binary)
│   ├── metadata.json             Text chunks with source filenames
│   └── embed_cache.json          Hash → vector cache for fast re-indexing
│
├── requirements.txt
├── .env.example
├── README.md
└── DOCS.md

Quick Start

Requirements:

Python 3.10, 3.11, or 3.12
Git
4 GB RAM minimum
Internet connection  (downloads models ~500 MB on first run)

Step 1 - Clone

git clone https://github.com/apdoolhamza/smartsupport-ai.git
cd smartsupport-ai

Step 2 - Install

python -m venv venv
source venv/bin/activate          # Windows: venv\Scripts\activate

pip install -r requirements.txt

Step 3 - Add your documents

Open the documents/ folder. Replace the sample files with your own content:

documents/
├── faq.txt        ← your questions and answers
└── products.txt   ← your product catalog

Write your FAQ in plain text using Q & A format:

Q: What are your opening hours?
A: We are open Monday to Friday, 9 AM to 6 PM,
   and Saturday 10 AM to 4 PM. Closed on Sundays.

Step 4 - Build the index

python scripts/build_index.py

This reads every file in documents/ and builds the FAISS index. You will see output like:

Found 2 document(s):
  - faq.txt
  - products.txt

Reading: faq.txt ...... OK (4,200 characters)
Reading: products.txt . OK (8,100 characters)

Building index...

Index built successfully.
Documents : 2  |  Chunks : 62  |  Vectors : 62

Step 5 - Configure

cp .env.example .env

Open .env and set your bot settings:

API_BASE_URL=https://your-server.com
BOT_NAME="Support AI"
PRIMARY_COLOR=#2563eb
SECONDARY_COLOR=#1d4ed8
WELCOME_MESSAGE=Hello! How can I help you today?
FALLBACK_MESSAGE="..."
WIDGET_POSITION=bottom-right
BUTTON_SIZE="56px"
API_KEYS=ss-abc123 (Optional)
...

Step 6 - Start the server

uvicorn backend.main:app --host 0.0.0.0 --port 8000 --reload

The server starts at http://localhost:8000.

Step 7 - Open in your browser

http://localhost:8000/health    ← confirm the server is running

Models download automatically on first run. After that they load from cache in seconds.

Step 8 - Embed on Your Website

Once your server is running, add one line to your website before </body>:

<script src="https://your-server.com/widget.js"></script>

A chat button appears on your website. Your customers can now get instant, accurate answers from your documents.

Note: If you set API Key in .env file, then you must set the key before the embed script in your website. You can also set the Bot config settings here:

<script>
window.__SS_CONFIG__ = {
    apiBase:        "https://your-server.com",
    botName:        "Support AI",
    primaryColor:   "#2563eb",
    secondaryColor: "#1d4ed8",
    welcomeMessage: "Hello! How can I help you today?",
    position:       "bottom-right",
    buttonSize:     "56px",
    apiKey:         "ss-abc123"    // must match EXACTLY what is in .env API_KEYS
};
</script>
 
<script src="https://your-server.com/widget.js"></script>

Customization

All customization is done through environment variables in .env or in your website before </body> closing tag. No code changes needed. Restart the server after editing.

Variable	What It Controls	Default
BOT_NAME	Name shown in the chat header	Support AI
PRIMARY_COLOR	Button color, user bubbles, and accent elements	#2563eb
SECONDARY_COLOR	Gradient secondary color	#1d4ed8
WELCOME_MESSAGE	First message sent when the widget opens	Hello! How can I help you today?
FALLBACK_MESSAGE	Reply when no relevant answer is found	See .env.example
WIDGET_POSITION	bottom-right or bottom-left	bottom-right
BUTTON_SIZE	Size of button	56px

API Reference

Interactive docs available at http://localhost:8000/docs

Method	Endpoint	Description
GET	/health	Server status and index statistics
POST	/api/chat	Submit a question, receive a full answer
POST	/api/chat/stream	Submit a question, receive a streamed response
GET	/widget.js	Chat widget with configuration pre-injected

Contributing

git checkout -b feat/your-feature
git commit -m "feat: describe your change"
git push origin feat/your-feature

---

MIT License · Built by apdoolhamza

If this saved you time, a star helps others find it.