Chatbot-for-Innovation 🤖✨

A lightweight, retrieval-augmented chatbot that lets you “talk” to innovation reports and datasets.

What it does

This notebook-driven project turns any PDF (and easily extensible data sources) into a conversational knowledge base:

1. Ingest – Extracts text with pdfplumber, chunks it using RecursiveCharacterTextSplitter, and embeds it with OpenAI models.
2. Store – Persists embeddings in a local Chroma vector database (./chroma_db) so you can pick up right where you left off.
3. Chat – Runs a ConversationalRetrievalChain (LangChain v0.1+) with memory, powered by GPT-4o/3.5 (ChatOpenAI), letting you ask follow-up questions naturally.
4. Share (optional) – Expose the bot to teammates with one command via pyngrok.

Typical use-cases include rapid policy analysis, literature deep dives, and interactive Q&A for hackathons or classroom demos.

---

Quick start

Prerequisites

• Python ≥ 3.10
• An OpenAI API key
• (Optional) Ngrok account & authtoken if you want a public URL
• Git + Jupyter

# 1. Clone
git clone https://github.com/<your-user>/Chatbot-for-Innovation.git
cd Chatbot-for-Innovation

# 2. Create & activate a virtual env (recommended)
python -m venv .venv
source .venv/bin/activate

# 3. Install dependencies
pip install -r requirements.txt

# 4. Configure your key
export OPENAI_API_KEY="sk-..."

# 5. Launch notebook
jupyter notebook Chatbot.ipynb

Example queries

• What are Canada's main innovation challenges? • How could Canada better compete with China and the U.S. in AI R&D? • Why does the report suggest embedding SR&ED consultants in innovation hubs?

Chatbot-for-Innovation/ ├── Chatbot.ipynb ← main notebook (ingestion + chat) ├── analysis.ipynb ← sample data-exploration notebook ├── chroma_db/ ← generated at runtime (persistent vectors) └── requirements.txt ← pinned dependencies