🦜 LangChain Exploration

A hands-on repository documenting my journey learning LangChain — from basic LLM integrations to building a functional terminal chatbot.

---

📁 Structure

Langchain/
├── Google-GenAI/       # LangChain with Google Gemini
├── Mistral-AI/         # LangChain with Mistral AI
└── lc-chatbot/         # Terminal chatbot project

---

🔵 Google-GenAI

Exploring LangChain with Google's Gemini model.

Model used: gemini-2.5-flash

from langchain_google_genai import ChatGoogleGenerativeAI

model = ChatGoogleGenerativeAI(model="gemini-2.5-flash")
response = model.invoke("Hello gemini")
print(response.text)

Setup:

pip install langchain-google-genai python-dotenv

Add your API key to .env:

GOOGLE_API_KEY=your_key_here

---

🟠 Mistral-AI

Exploring LangChain with Mistral AI's chat model.

Model used: mistral-small-latest

from langchain_mistralai.chat_models import ChatMistralAI

model = ChatMistralAI(model="mistral-small-latest")
response = model.invoke("Hello Mistral")
print(response.text)

Setup:

pip install langchain-mistralai python-dotenv

Add your API key to .env:

MISTRAL_API_KEY=your_key_here

---

🤖 lc-chatbot

A terminal-based conversational chatbot built with LangChain and Google Gemini. Maintains full chat history across the session so the model remembers previous messages.

Model used: gemini-2.5-flash

from langchain_google_genai import ChatGoogleGenerativeAI
from langchain.messages import HumanMessage, AIMessage

model = ChatGoogleGenerativeAI(model="gemini-2.5-flash")
messages = []

while True:
    userInput = input("Enter Prompt ......")
    messages.append(HumanMessage(userInput))
    response = model.invoke(messages)
    messages.append(AIMessage(response.content))
    print(response.text)

Run it:

cd lc-chatbot
pip install -r requirements.txt
python main.py

🟣 AI-Tools

Exploring LangChain Tools, LLM hallucinations, and tool calling workflows using Mistral.

This module demonstrates how to overcome one of the biggest limitations of LLMs — lack of real-time awareness — by integrating external tools.

---

❗ Problem: LLM Hallucination

LLMs like Mistral:

• Do not know real-time data
• Can generate incorrect or guessed answers

Example:

model.invoke("What is today's date?")

➡️ May return a wrong date

---

✅ Solution: Tool Integration

We define a tool to fetch real-time data:

from langchain.tools import tool
from datetime import date

@tool
def getCurrentDate():
    """Returns current date"""
    return str(date.today())

---

⚙️ Tool Calling with Mistral

Bind tool to model:

model = ChatMistralAI(model="mistral-small").bind_tools([getCurrentDate])

---

🔁 Execution Flow

response = model.invoke("today's date ?")

tool_result = getCurrentDate.invoke(response.tool_calls[0]["args"])

final_response = model.invoke(
    ["Human:- Today's Date", "Tool result := " + tool_result]
)

print(final_response.text)

---

🔄 How It Works

1. User asks a question
2. LLM decides to call a tool
3. Tool executes (getCurrentDate)
4. Result is passed back to LLM
5. LLM generates accurate final answer

---

🎯 Key Learnings

• Tools solve hallucination for dynamic data
• LLM + Tools = more reliable systems
• This is the foundation of AI agents

---

📌 Future Scope

• Automate tool execution loop
• Add multiple tools
• Build full agent using AgentExecutor

---

⚙️ General Setup

1. Clone the repo:

git clone https://github.com/amitava-code/Langchain.git
cd Langchain

2. Create and activate a virtual environment:

python -m venv venv
source venv/Scripts/activate  # Git Bash on Windows

3. Install dependencies and add your API keys to a .env file in the relevant folder.

---

🛠️ Tech Stack

• LangChain
• Google Gemini
• Mistral AI
• Python 3.11+