Repository files navigation

English Doc中文文档

ChatWeb can crawl any webpage or extract text from PDF, DOCX, TXT files, and generate an embedded summary.It can also answer your questions based on the content of the text.It is implemented using the chatAPI and embeddingAPI based on gpt3.5, as well as a vector database.

The basic principle is similar to existing projects such as chatPDF and automated customer service AI.

Crawl web pagesExtract text contentUse GPT3.5's embedding API to generate vectors for each paragraphCalculate the similarity score between each paragraph's vector and the entire text's vector to generate a summaryStore the vector-text mapping in a vector databaseGenerate keywords from user inputGenerate a vector from the keywordsUse the vector database to perform a nearest neighbor search and return a list of the most similar textsUse GPT3.5's chat API to design a prompt that answers the user's question based on the most similar texts in the list.The idea is to extract relevant content from a large amount of text and then answer questions based on that content, which can achieve a similar effect to breaking through token limits.

An improvement was made to generate vectors based on keywords rather than the user's question, which increases the accuracy of searching for relevant texts.

• Install Python3
• Download this repository by runninggit clone https://github.com/SkywalkerDarren/chatWeb.git
• Navigate to the directory by runningcd chatWeb
• Copyconfig.example.json toconfig.json
• Editconfig.json and setopen_ai_key to your OpenAI API key
• Install dependencies by runningpip3 install -r requirements.txt
• Start the application by runningpython3 main.py

if you prefer, you can also run this project using docker:

• build the container usingdocker-compose build (only needed once when you are not planning to contibute to this repo)
• copyconfig.example.json toconfig.json and set all the needed stuff. The example config is already fine for running with docker, no need to change anything there, if you don't have the OPEN_AI_KEY in your env variables you can set it here too, or later if you run this app.
• run the container: `docker-compose up"
• open the application in browser:http://localhost:7860

• Editconfig.json, setlanguage toEnglish or other language

• Editconfig.json and setmode toconsole,api, orwebui to choose the startup mode.
• Inconsole mode, type/help to view commands.
• Inapi mode, an API service can be provided to the outside world.api_port andapi_host can be set inconfig.json.
• Inwebui mode, a web user interface service can be provided.webui_port can be set inconfig.json, defaulting tohttp://127.0.0.1:7860.

• Editconfig.json and setuse_stream totrue.

• Editconfig.json and settemperature to a value between 0 and 1.
• The smaller the value, the more conservative and stable the response will be. The larger the value, the more daring the response may be, possibly resulting in "hallucinations."

• Editconfig.json and addopen_ai_proxy for your proxy address, for example:

"open_ai_proxy": {  "http": "socks5://127.0.0.1:1081",  "https": "socks5://127.0.0.1:1081"}

• Editconfig.json and setuse_postgres totrue.
• Install PostgreSQL.
  • The default SQL address ispostgresql://localhost:5432/mydb, or you can set it inconfig.json.
• Install the pgvector plugin.

Compile and install the extension (support Postgres 11+).

git clone --branch v0.4.0 https://github.com/pgvector/pgvector.gitcd pgvectormakemake install# may need sudo

Then load it in the database you want to use it in

CREATE EXTENSION vector;

• Install dependency with pip:pip3 install psycopg2

Please enter the link to the article or the file path of the PDF/TXT/DOCX document: https://gutenberg.ca/ebooks/hemingwaye-oldmanandthesea/hemingwaye-oldmanandthesea-00-e.htmlPlease wait for 10 seconds until the webpage finishes loading.The article has been retrieved, and the number of text fragments is: 663...=====================================Query fragments used tokens: 7219, cost:$0.0028876Query fragments used tokens: 7250, cost:$0.0029000000000000002Query fragments used tokens: 7188, cost:$0.0028752Query fragments used tokens: 7177, cost:$0.0028708Query fragments used tokens: 2378, cost:$0.0009512000000000001Embeddings have been created with 663 embeddings, using 31212 tokens, costing$0.0124848The embeddings have been saved.=====================================Please enter your query (/help to view commands):

• Support for pdf/txt/docx files
• Support for in-memory storage without a database (faiss)
• Support for Stream
• Support for API
• Support for proxies
• Add Colab support
• Add language support
• Support for temperature
• Support for webui
• Other features that have not been thought of yet