This project has ceased active development. PRs will still be accepted.

Side Project Idea Generator

Helps software development students generate project ideas based on their skills and interests. The final version of this product will run on the following stack:

• Python/FastAPI for prompt-management and handling to LLM provider server
• React front end with Mantine component library

Requirements

• At least one of the following:

  1. Fast computer with > 8GB of RAM to run the offline LLM inference server

     or

  2. An API key with one of the supported LLM services (see below)

• Python version 3.11 or greater installed

LLM Setup

This package requires that you have an LLM inference server setup. Jan is a recommended desktop GUI application that runs an LLM inference on your computer, no internet connection needed. Ollama is another popular option, but its API is not directly compatible with the OpenAI python library and is therefore not supported at this time.

Jan instructions

⚠️ The default model in JanAi, phi-2-3b, requires 8GB of RAM and closing all your other apps, otherwise your computer will grind to a halt.

Settings -> Model: choose Phi-2 3B Q8 (requires 8GB of RAM, closed applications, and a fast computer)

Settings -> Advanced: Enable API Server

OpenAI / Anyscale / TogetherAi API

You can run this package using OpenAI's, AnyScale's, or TogetherAi's servers, but you will need to obtain an API key and add it in your .env file. See Step 6 below.

You can get $25 in free credits with TogetherAi, and $10 in free credits from Anyscale when you sign up for their services. No payment details needed.

Installation

1. Install Python version >= 3.11
2. git clone the repo and cd into the project directory
3. Set up a virtual environment

python3 -m venv .venv

4. Activate the virtual environment (macOS/Linux)

source .venv/bin/activate

5. Install the requirements

pip install -r requirements.txt

6. (Optional) Rename your .env file if you plan on using cloud LLMs

mv .env.example .env

7. (Optional) Add your OpenAI/AnyScale keys to your .env file
8. (Optional) Open main_cli.py or main.py and switch your LLM provider from Jan to another listed in the config.toml file.

config = Config("JanAi") # Switch "JanAi" to "OpenAI" etc

8. (Optional) To use a terminal command line, execute the CLI entrypoint

python3 main_cli.py

9. Start the FastAPI endpoint

uvicorn main:app --reload

10. Start the web service

cd frontend

Install web dependencies

npm install

Start the web server

npm run dev

API Endpoints

Use a tool like Postman to test the endpoints.

http://127.0.0.1:8000/prompt

If you are developing the frontend and need a mocked response to avoid pinging an LLM inference server, use the test endpoint.

http://127.0.0.1:8000/test

JSON input:

{
    "unknown_tech": ["Backend", "databases", "Java"],
    "known_tech": ["Javascript", "React", "Typescript", "Vue"],
    "topics": ["Dancing", "Cooking"]
}

Example JSON response:

{
    "project_title": "Dance Recipe App",
    "description": "An app that combines dancing and cooking by providing users with fun dance routines while preparing recipes, creating an interactive and enjoyable cooking experience.",
    "technical_requirements": [
        "Develop the app using React and Typescript for the frontend to ensure a robust and efficient user interface.",
        "Incorporate Vue.js for interactive and visually engaging dance routine display and user interaction.",
        "Utilize JavaScript for implementing audio playback functionality and dance routine synchronization with recipe steps."
    ],
    "user_stories": [
        "As a user, I can browse through a variety of recipes and select one to prepare.",
        "As a user, I can view and follow along with a dance routine that complements the cooking process for a selected recipe.",
        "As a user, I can pause, replay, or adjust the speed of the dance routine to match my preference and cooking pace.",
        "As a user, I can see the ingredients and cooking steps for a recipe while simultaneously watching and following the dance routine.",
        "As a user, I can track my progress and save my favorite recipes and dance routines for future use."
    ]
}

Todo list

LLM Service

• Build skeletal code for abstracting LLM providers and clients from OpenAI implementation
• Add multiple LLM API urls and models
• Add Pydantic models and validation
• Create toml file to store endpoint configs
• Refactor main_cli.py to emulate a FastAPI entrypoint
• Add Pydantic error handling
• Add retry logic for failed validation
• Convert main_cli.py to FastAPI main.py
• Add streaming support
• Add a mocking capability to ease frontend development
• Add rate limiting middleware

Database

• Create caching schema

Frontend

• Create blank React project with Vite
• Build barebones app shell with Mantine
• Create form elements with Mantine/Forms
• Create list of technologies/topics to choose from
• Implement fetch to Uvicorn API

Infrastructure

• Dockerize the project
• Deploy to Cloudflare to sxflynn.net subdomain

Contributing

Anyone is welcome to contribute bug fixes and ideas as an issue. Unless it's a quick fix or a documentation enhancement, please report your idea as an issue before submitting a PR.