README.md

---
license: mit
title: Freeekyyy-chatBot
sdk: streamlit
sdk_version: 1.44.1
---
# 🤖 Freeekyyy ChatBot

**Freeekyyy** is an *over-the-top*, emotional AI chatbot that FREAKS OUT (in Markdown!) on any topic you provide.  
It uses [LangChain](https://github.com/langchain-ai/langchain) + [OpenRouter](https://openrouter.ai) to generate expressive, explosive Markdown responses — perfect for dramatic, chaotic, and wildly informative outputs.

> 🔥 Now powered with a **RAG (Retrieval-Augmented Generation) pipeline** to respond using your own PDFs and documents!

Check it out live 👉 [MKCL/Freeekyyy-chatBot on Hugging Face 🤯](https://huggingface.co/spaces/MKCL/Freeekyyy-chatBot)

---

## 🧠 How It Works

- Uses `LangChain`'s `ChatPromptTemplate` to inject emotional few-shot prompts.
- Connects to **DeepSeek-R1-Zero** via [OpenRouter](https://openrouter.ai).
- Uses **vector search** (via `ChromaDB`) and **HuggingFace embeddings** for document retrieval (RAG).
- Outputs responses in beautiful **Markdown (.md)** format.
- Works as a **Streamlit app** or a **FastAPI backend**.

---

## 🔍 Retrieval-Augmented Generation (RAG)

The chatbot now includes a smart document processing pipeline:

1. **Document Ingestion**: Parses your uploaded PDF files.
2. **Chunking**: Splits them into overlapping text chunks.
3. **Embeddings**: Generates vector embeddings using `BAAI/bge-small-en`.
4. **Vector Store**: Stores chunks in `ChromaDB`.
5. **Context Injection**: Relevant chunks are inserted into the LLM prompt for context-aware responses!

---

## 🖥️ Streamlit Integration

To display Markdown output in Streamlit:

```python
import streamlit as st

# Assuming `md_output` contains your model's response
st.markdown(md_output, unsafe_allow_html=True)
```

---

## 🚀 Installation

### Option 1: Using `uv`
```bash
uv pip install -r requirements.txt
```

### Option 2: Using regular pip
```bash
pip install -r requirements.txt
```

---

## 📦 Requirements

```
langchain
langchain-community
langchain-openai
openai
chromadb
python-dotenv
huggingface_hub
sentence-transformers
streamlit
uvicorn
fastapi
```

---

## 🛠️ Environment Variables

Create a `.env` file in the root directory:

```
OPENROUTER_API_KEY=your_openrouter_key_here
HUGGINGFACE_API_KEY=your_huggingface_key_here
```

---

## 🧪 Example Prompt Structure

```python
from langchain.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You're an extremely emotional AI. Always freak out in Markdown."),
    ("user", "Topic: Volcanoes")
])
```

---

## 🔗 RAG Query with Vector Search

```python
# Sample retrieval pipeline
relevant_chunks = db.similarity_search(query, k=4)
context = "\n\n".join([doc.page_content for doc in relevant_chunks])

final_prompt = f"""
You are an emotional assistant. Respond dramatically using Markdown.

Context:
{context}

Question:
{query}
"""
```

---

## 🧑‍💻 Want to Use as an API?

Run your backend like this:

```bash
uvicorn main:app --reload
```

---

## 📎 License

MIT — go freak out and teach some AI emotions! 🤯❤️🔥