BigQuery Chatbot is an intelligent conversational AI that interacts with Google BigQuery to fetch and analyze data. It integrates LangChain, Streamlit, and LangSmith for real-time query execution, schema inspection, and chatbot monitoring.

✅ Conversational AI: Uses Google Generative AI (Gemini) for intelligent responses.
✅ BigQuery Integration: Fetches schemas and executes queries in real-time.
✅ Session-based Memory: Maintains persistent conversation history.
✅ LangSmith Integration: Tracks and evaluates chatbot performance.
✅ Streamlit UI: User-friendly web interface for interacting with BigQuery.

git clone https://github.com/espin086/BigQueryChat
cd bigquery_chatbot

python -m venv venv
source venv/bin/activate  # On macOS/Linux
venv\Scripts\activate      # On Windows

pip install -r requirements.txt

To configure the chatbot, create a .env file by copying the provided .env.template:

cp .env.template .env

Then, open .env and update the values with your Google Cloud and LangSmith API credentials:

GOOGLE_API_KEY=your_api_key_here
GOOGLE_APPLICATION_CREDENTIALS=/path/to/your_gcp_json_key.json
PROJECT_ID=your_bigquery_project_id
DATASET_ID=your_bigquery_dataset_name
LANGCHAIN_PROJECT=your-langsmith-project
LANGCHAIN_API_KEY=your-langsmith-api-key

Note: Make sure to provide the correct Google Cloud service account JSON key to authenticate BigQuery access.

streamlit run streamlit_app.py

Chatbot can retrieve the schema of a BigQuery table. Example prompt:

You can execute SQL queries using natural language prompts:

bigquery_chatbot/
│── src/
│   ├── agent.py              # Defines LangChain agent
│   ├── tools.py              # Tools for fetching schema and executing queries
│   ├── bigquery_manager.py   # Handles BigQuery interactions
│   ├── sql_lite_handler.py   # Manages SQLite database for chat history
│   ├── format_message.py     # Formats the LLM response for display
│   ├── utils/
│       ├── config.py         # Database configurations
│── streamlit_app.py          # Streamlit UI
│── requirements.txt          # Dependencies
│── README.md                 # Documentation

LangSmith is used to trace, monitor, and evaluate chatbot performance in real-time.

1. Go to LangSmith ➡️ LangSmith Platform

2. Create a Project and Generate an API Key (If you haven't already).

3. Update Your .env File with the LangSmith details:

   LANGCHAIN_PROJECT=your-langsmith-project-name
   LANGCHAIN_API_KEY=your-langsmith-api-key

Add the following code snippet to ensure LangSmith is tracking all interactions:

from langchain.callbacks.tracers import LangChainTracer
from dotenv import load_dotenv
import os

# Load API keys from environment variables
load_dotenv()
langsmith_tracer = LangChainTracer()

llm = ChatGoogleGenerativeAI(
    model="gemini-1.5-pro",
    temperature=0,
    max_tokens=None,
    callbacks=[langsmith_tracer],
)

Now, all chatbot interactions will be logged in LangSmith!

Once the chatbot is running, you can track its performance via the LangSmith Dashboard.