https://github.com/perrys25/SearchGPT

https://github.com/perrys25/SearchGPT

google-search gpt-35-turbo

Last synced: about 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/perrys25/SearchGPT
• Owner: perrys25
• License: mit
• Created: 2023-02-19T22:57:11.000Z (almost 2 years ago)
• Default Branch: master
• Last Pushed: 2023-03-20T16:28:01.000Z (almost 2 years ago)
• Last Synced: 2024-07-31T08:18:25.018Z (5 months ago)
• Topics: google-search, gpt-35-turbo
• Language: Python
• Homepage: https://searchgpt.perrysahnow.com/
• Size: 29.3 KB
• Stars: 33
• Watchers: 2
• Forks: 8
• Open Issues: 0
• Metadata Files:
  • Readme: readme.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# SearchGPT

Bringing the Internet, and Chat AI one step closer

## What is SearchGPT?

SearchGPT is a library for using [OpenAI](https://openai.com/)'s GPT-3 Language Model & Google Search to power a Question-Answering Chatbot similar to that of Bing's assistant that can do external research as needed to help answer questions

## How can I try it?

**Option 1:** Visit the hosted demo @ [searchgpt.perrysahnow.com](https://searchgpt.perrysahnow.com/)

**Option 2:** [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/gist/perrys25/195fd36b6986025e7711b55189cae9b9/searchgpt-demo.ipynb)

**Option 3:**

1. Clone `SearchGPT`'s git Repo

2. Using Python `3.9`

3. Setup a python venv (May vary based on your system / IDE)

4. Run `pip install -r requirements.txt` (Make sure you install chrome!)

5. Open `examples`

6. Run `py cli.py`

## How does the API Work?

### ChatBot

The main API is a `ChatBot` class of which you can create an instance, and ask questions

An example of creating a bot named "Steve" who thought he was voice activated would be the following:

```py

bot = ChatBot("Steven", type="Voice activated search assist chatbot", commands=[SearchCommand()])

bot.ask("What's the weather like right now?")

```

### Commands

Commands help the bot perform actions. two example command are the `SearchCommand` used for searching google, or the FollowUP command inside the `cli.py` UI used for asking secondary questions when needed

An example for a fictional home assistant controller would be:

```py

class LightsCommand(Command):

    def getName(cls):

        return "Lights"

    def getDescription(cls):

        return """

           If the user has asked to turn out the lights in a specific room (which must be decided on beforehand, possibly with a followup), use the "[Lights] Off RoomName" Command

           It will always be followed by a "[Lights Response]" command either containing "Success", or an encountered error while disabling lights

           """

    def execute(self, chatbot, data, question):

        if data.split(" ")[0] == "Off":

            # Turn off lights

            return "Success"

        return "Failure, invalid command"

```