Updated Langchain for GenAI Pet project v2 Download - Updated Langchain for GenAI Pet project v2 Source code download

Updated Langchain for GenAI Pet project v2

Other source code

1.0.0

Download

Building Chatbot Using Paid And Open Source LLM's using Langchain And Ollama

This End-to-end project that simplifies the lifecycle of Large Language Model (LLM) applications, including development, productionization, and deployment. The repository contains folders and files as components of the project, such as agents, APIs, chains, chatbots, GROQ (a query language), Hugging Face models, ObjectBox (an embedded database), OpenAI models, and Retrieval-Augmented Generation (RAG). ) models. The project aims to provide a comprehensive solution for working with LLMs, covering data handling, chatbot development, and integration with various tools and frameworks.

Langchain Project

Chatbot test runs

String output parser example

Building a Chatbot Using Paid and Open Source LLMs with Langchain and Ollama

Project Overview

This project demonstrates the creation of a chatbot using both paid and open source large-scale language models (LLMs) via Langchain and Ollama. It covers the entire life cycle of LLM applications, including development, production and deployment.

Information security:

Yeah. All APIs are secret, with restricted user access and anonymized .gitignore and another layer of security through Github Secret Environments (Sorry! you'll have to use your own!)

Minimum Project Components

Agents
APIs / FastAPI
Chains
Chatbots
GROQ (query language)
Hugging Face22 Models
ObjectBox (embedded database)
OpenAI models
Meta Llama2 Models (open source)
Recovery Augmented Generation (RAG) Models

Configuration and Installation

1. Environment Configuration

conda create -p venv python==3.10 -y
venv S cripts a ctivate
pip install -r requirements.txt

2. Configuration

Configure the necessary API keys and settings in the app.py file.

Use

1. Chatbot Execution

python app.py

2. Access to the Web Interface

Open a web browser and navigate to (or the assigned progressive port on your computer). mine is that:

http://localhost:8502

Let's configure app.py and run the session on local host

We create the file by calling the 3 main chained functions of our base chatbot.
We load requirements for modules and libraries:
We run our app.py
We finish configuring app.py and run

It returns the route and opens the host in the test browser alt text

Local URL: http://localhost:8502
Network URL: http://192.168.100.30:8502

Let's check the page:
Let's review the project in the Langchain framework:
Let's look at the structure of the output as we define it:
We look at the StrOutputParser in detail:
We will ask you about “providing python code that exchanges two values” to evaluate cost, tokenization and delay time to resolve the question:

*proveeme de un código python que swapee 2 valores.* :

alt text

 # Definir los dos valores a intercambiar
a = 5
b = 10

print ( "Valores originales:" )
print ( "a =" , a )
print ( "b =" , b )

# Intercambiar los valores
temp = a
a = b
b = temp

print ( " n Valores intercambiados:" )
print ( "a =" , a )
print ( "b =" , b )

Finally we generate a common query:
“question": "When to start brushing puppies' teeth?"

Let's configure a Local version with Ollama

We will load Llamama specifically by downloading Llama2 to our computer to run it from my local environment
- I open a cmd
```
ollama run llama2
```
- we download llama2
- We will consult the same last 2 questions to evaluate the level of response, coherence and consistency with Llama2 compared to that generated with gpt3.5 turbo
- We look at the responses in the test browser:
- Let's review langsmith on Langchain to review query tracing, latency, tokenization and the entire process.
- “provide me a python code to swap 2 numbers”
- How we see the process time and generation of the response
- It was much higher, and tokenization quickly skyrocketed to 388 tokens.
- The size and length of the answer is generated by the question, but much more. And the level of information and completeness is also high, which means that the OpenAI model I try to respond with brevity and precision, without going into semantic details while the model works.
- Lama 2 was much more verbal or verbose and with an impact associated with the local processing time that depends on the computer on which the model is running and the intelligence behind the response generated by the model itself.