https://github.com/natserract/natserract-ai

Using Doc2Vec, Langchain and OpenAI to chat with Natserract blog https://engineering-natserract.vercel.app/
https://github.com/natserract/natserract-ai

chatbot doc2vec gpt4 langchain natserract openai pgvector qatool

Last synced: 5 days ago
JSON representation

Using Doc2Vec, Langchain and OpenAI to chat with Natserract blog https://engineering-natserract.vercel.app/

• Host: GitHub
• URL: https://github.com/natserract/natserract-ai
• Owner: natserract
• Created: 2023-11-09T02:52:11.000Z (about 1 year ago)
• Default Branch: master
• Last Pushed: 2023-11-16T08:38:37.000Z (about 1 year ago)
• Last Synced: 2024-04-28T06:27:30.218Z (8 months ago)
• Topics: chatbot, doc2vec, gpt4, langchain, natserract, openai, pgvector, qatool
• Language: Python
• Homepage:
• Size: 7.79 MB
• Stars: 4
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Natserract AI

This repository contains code that demonstrates how to build AI assistant using Langchain, integrating GPT-4 from OpenAI. The assistant can handle question-answering (QA), provide various tools, similarity search with Doc2Vec approach, to provide answers to user queries based on the provided documents.

Throughout this journey, i use PostgreSQL as the main Database and PGVector extension to store the embeddings.

## Setup

Before running the script, you need to set up the required credentials and install the necessary libraries.

### Install Required Libraries

You can install the required libraries using poetry. Run the following command in your terminal or command prompt:

```sh

poetry install

```

### Install Spacy

```sh

python -m spacy download en_core_web_sm

```

### Setup API Keys

The script uses the OpenAI API key. You need to set these API keys as environment variables in your system. Replace OPENAI_API_KEY with your actual API keys.

### Setup Database

- Postgres 15

- Enable the extension

  ```sql 

  CREATE EXTENSION vector;

  ```

## Running

```sh

poetry shell

poetry run python main.py

```

## Process

![](process.png)

## Demo

https://github.com/natserract/natserract-ai/assets/31182611/8b4be411-1fd4-4c9f-89cc-eb25670f7ead

## Custom Datasets

Create `_datasets` directory and place all markdown documents in it.

## Performance Considerations

If you need to perform this operation frequently and especially if the set of word vectors is large, it may be practical to use a database or a data store optimized for vector operations. These data stores can persist your word vectors and provide efficient similarity search functionality:

- FAISS by Facebook AI Research is a library for efficient similarity search and clustering of dense vectors.

- Elasticsearch has plugins like elasticsearch-vector-scoring to handle vector similarity.

- Annoy (Approximate Nearest Neighbors Oh Yeah) is a C++ library with Python bindings to search for points in space that are close to a given query point.

Using such systems can significantly speed up the similarity search process.