V-Chat: AI Chatbot for Sustainability Platform

V-Chat is an AI-powered conversational agent developed for integration into a sustainability-focused web platform. The chatbot is designed to support user interaction within the agriculture and environmental sustainability domain. It is built to provide contextual assistance, information retrieval, and guidance for platform users.

This project uses LangChain, Pinecone, and Gemini 2.0 Flash to deliver relevant answers based only on the platform's verified data.

The chatbot:

• Takes a user's question.
• Searches the vector database (Pinecone) for relevant information from the villam_hub_knowledge_base.md document.
• Uses AI model (Google's Gemini 2.0 Flash) to understand the question and the retrieved information.
• Generates a concise and helpful answer.

GitHub repository: https://github.com/DataxEnv/v-chat-sustainability-chatbot

Features

• AI-Powered Responses: uses Google's Gemini 2.0 Flash model for intelligent answers.
• Context-Aware: Retrieves relevant information from villam_hub_knowledge_base.md using Pinecone vector search.
• Conversational Memory: Remembers previous parts of the conversation to provide more relevant follow-up answers.

Tech Stack

• Python: The primary programming language.
• Langchain: A framework to simplify building applications with Large Language Models (LLMs).
• Google Generative AI (Gemini 2.0 Flash): The LLM used for understanding and generating text.
• GoogleGenerativeAIEmbeddings: For converting text into numerical representations (embeddings) that AI can understand.
• Pinecone: A vector database used to store and efficiently search through the embeddings of our knowledge base.
• Dotenv: To manage sensitive information like API keys.

Purpose

This repo contains the backend logic and testing environment for V-Chat. It is meant for the Internal team members.

Project Structure

File/Folder	Description
.gitignore	Hides the .env containing API keys
.env.example	API keys template
requirements.txt	Lists all Python libraries needed to run the chatbot
villam_hub_knowledge_base.md	Cleaned demo dataset containing Villam Hub knowledge
vchat_pipeline.ipynb	Main chatbot logic for exploration and debugging: embeds queries, retrieves data, and generates responses
vchat.py	Main chatbot logic for integration
vchat_pinecone.ipynb	Notebook to load and upsert the dataset into Pinecone
test_chatbot.py	Notebook to test chatbot responses interactively

Setup Instructions

1. Clone the Repo

git clone https://github.com/olamide-analyst/Villam-chatbot v-chat
cd V-Chat

2. Add Your Environment Variables (.env file)

Add API keys to .env.example file and rename to .env :

GOOGLE_API_KEY=google_generative_ai_key
PINECONE_API_KEY=pinecone_key

.env is already excluded from version control via .gitignore.

3. Create and Activate a Virtual Environment (recommended)

macOS/Linux (zsh/bash):

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip

Windows (PowerShell):

py -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip

Note:

• Your prompt should show (.venv) when activated.
• Run deactivate to exit the environment.

4. Install Required Libraries

Install all the Python packages listed in requirements.txt inside the activated virtual environment:

pip install -r requirements.txt

If you plan to use the notebooks in this repo with this environment, add the kernel to Jupyter:

python -m ipykernel install --user --name villam-chatbot-venv --display-name "Python (.venv)"

How to use

The project has two main Python scripts:

• vchat_pinecone.ipynb: This script processes the dataset villam_hub_knowledge_base.md, converts its content into embeddings/vectors (numerical representations), and stores(upsert) them in the Pinecone vector database. You only need to run this once (or whenever the dataset document changes).

• vchat_pipeline.ipynb or vchat.py: These runs the chatbot engine logic (retrieval + Gemini 2.0 generation).

Before running examples below, make sure your virtual environment is active (source .venv/bin/activate).

Step 1: Upsert the Dataset to pinecone

Run vchat_pinecone.ipynb to upload the contents of villam_hub_knowledge_base.md to Pinecone.

Step 2: Test the Chatbot

Use either:

• test_chatbot.py for interactive testing

streamlit run test_chatbot.py

• Or vchat_pipeline.ipynb to debug the full flow end-to-end

Backend Integration

For developers:

• All chatbot logic lives in generate_response() inside vchat.py

This is ready to be integrated with a Flask, FastAPI, or Streamlit frontend.

Notes

• The chatbot uses Retrieval-Augmented Generation (RAG).

• Gemini 2.0 Flash is rate-limited on the free tier:

  • 15 requests/minute
  • 200 requests/day

Contributing

This project is maintained by the Villam Hub team.

References

• YouTube Gemini + LangChain Tutorial: https://youtu.be/DFjuV2YBoe4?si=2ND-frk2_Wjfv9FF