GSAM ✨ GenericSuite App Maker

AI tool to enhance the software development ideation and AI models, providers and features selection and test.

Introduction

GSAM is a tool to help on the software development process for any Application. It allows to generate code, images, video or answers from a text prompt, and kick start code to be used with the GenericSuite library.

Key Features

• Answer question with LLM inference, using Meta Llama models, Together.ai, HuggingFace, Groq, Ollama, Nvidia NIMs, and OpenAI.
• Image Generation: using HuggingFace and the Flux or OpenAI Dall-E models.
• Video Generation: using Rhymes AI Allegro model.
• Galleries to show the generated images and videos.
• Ability to change the Provider and Model used for all the LLM Inferences, image and video generations.
• Suggestions to generate App ideas, and the hability to customize the suggestion generation prompt.
• Code Generation: suggest the JSON configuration files and Langchain Tools Python code from an App description to be used with the GenericSuite library.
• Use LlamaIndex to generate code and JSON files using vectorized data instead of send all the attachments to the LLM.
• Store each user interaction (question, answer, image, video, code) in a MongoDB database, and retrieve it later.
• Database Management: import and export data from MongoDB to JSON files.
• Prompt Engineering: there's an option to allow the prompts/questions optimization to take more advantage from the Model's capabilities.
• Naming: generate name ideas for the App.
• App Structure: generate the App description and database table structures.
• App Presentation: generate PowerPoint presentation for the App, including the content, speaker notes, and image generation prompts.

Technology Used

• Meta Llama models: Llama 3.2 3B, Llama 3.1 8B, 70B, and 405B
• Together.ai
• Huggingface Inference API
• Flux.1 image generation model
• Rhymes Allegro video generation model
• LlamaIndex framework.
• StreamLit
• MongoDB Atlas
• Python 3.10

Getting Started

Prerequisites

• Python 3.10 or higher
• Git
• Make: Mac | Windows

Installation

Clone the repository:

git clone https://github.com/tomkat-cr/genericsuite-app-maker.git

Navigate to the project directory:

cd genericsuite-app-maker

Create the .env file

Create a .env file in the root directory of the project:

# You can copy the .env.example file in the root directory of the project
cp .env.example .env

The .env file should have the following content:

PYTHON_VERSION=3.10
#
# Together AI
TOGETHER_AI_API_KEY=
# HuggingFace
HUGGINGFACE_API_KEY=
# RHYMES parameters
RHYMES_ALLEGRO_API_KEY=
RHYMES_ARIA_API_KEY=
# OpenAI
OPENAI_API_KEY=
# Groq
GROQ_API_KEY=
# Ollama
OLLAMA_BASE_URL=localhost:11434
# Nvidia
NVIDIA_API_KEY=
#
# Database parameters
DB_TYPE=mongodb
# DB_TYPE=json
#
# MongoDB database parameters
MONGODB_URI=mongodb+srv://<user>:<password>@<cluster>.mongodb.net
MONGODB_DB_NAME=gsam_db
MONGODB_COLLECTION_NAME=conversations
#
# JSON database parameters
# JSON_DB_PATH=./db/conversations.json

Replace TOGETHER_AI_API_KEY and other access tokens with your actual Together.ai API key, OpenAI, Huggingface, Groq, Nvidia, and Rhymes API keys, respectively.

To use a MongoDB database, comment out DB_TYPE=json, uncomment # DB_TYPE=mongodb, and replace YOUR_MONGODB_URI, YOUR_MONGODB_DB_NAME, and YOUR_MONGODB_COLLECTION_NAME with your actual MongoDB URI, database name, and collection name, respectively.

Run the Application

# With Make
make run

# Without Make
sh scripts/run_app.sh run

Usage

Go to your favorite Browser and open the URL provided by the application.

• Locally:
  http://localhost:8501

• Official App:
  https://gsam-app.streamlit.app/

Prompt Suggestions

• The Prompt Suggestions under the title can be generated from AI using the Suggestions Prompt pull-down section. Enter the Prompt in the text box and click the Generate Suggestions button. Click on Reset Prompt to set the default prompt.
• Any suggestion text can be copied to the Question box by clicking on it.

Models Selection

• The LLM Chat, Image and Video generarion Providers and Models can be selected using the Models Selection pull-down section.

Text-to-Text Generation

• Enter your text prompt in the provided text box or select one of the suggested prompts.
• Check the Enhance prompt checkbox to allow the LLM to optimize the prompt to take more advantage from the Model's capabilities.
• Select the Main tab and click the Answer Question button.
• The answer will appear in below the queston box.
• Click the Use Response as Prompt button to use the answer as the prompt for the next question.
• All questions and answers are available in the side menu.

Text-to-Image Generation

• Enter your text prompt in the provided text box or select one of the suggested prompts.
• Check the Enhance prompt checkbox to allow the LLM to optimize the prompt.
• Select the Main tab and click the Generate Image button.
• Sit back and watch as GSAM transforms your text into a high-quality image.
• After a few seconds, the image will appear.
• All images are available in the side menu and can be viewed in the gallery clicking the Image Gallery button.

Text-to-Video Generation

• Enter your text prompt in the provided text box or select one of the suggested prompts.
• Check the Enhance prompt checkbox to allow the LLM to optimize the prompt.
• Select the Main tab and click the Generate Video button.
• Sit back and watch as GSAM transforms your text into a high-quality video.
• After 2+ minutes, the video will appear in the video container.
• All videos are available in the side menu and can be viewed in the gallery clicking the Video Gallery button.

App ideation

Click the App ideation tab to have access to the app ideation page. This page allows you to specify the application name, description, and other elements to generate naing ideas, app extended description and database structure, and presentation.

• Generate App Names: In the Application Ideation Form section, fill in the required fields (Application name, Subtitle, Summary and App Type) and click the Generate App Names button at the bottom to generate ideas on how to name you application.

• Generate App Structure: In the Application Ideation Form section, fill in the required fields (Application name, Subtitle, Summary and App Type) and click the Generate App Structure button at the bottom to generate ideas on how to descroibe and structure your application.

• Generate Presentation: In the Application Ideation Form section, fill in the all the form fields and click the Generate Presentation button at the bottom to generate the slides structure and create the PowerPoint file.

Code Generation

This option allows the JSON configuration files and Langchain Tools Python code generation from an App description to be used with the GenericSuite library.

• Enter your text prompt in the provided text box or select one of the suggested prompts.
• Check the Use Embeddings checkbox to use LlamaIndex in the code and JSON files generate, using vectorized data instead of send all the attachments to the LLM.
• Check the Enhance prompt checkbox to allow the LLM to optimize the prompt.
• Select the Code Generation tab and click the Generate Config & Tools Code button.
• Sit back and watch as GSAM transforms your text into a high-quality code.
• After a few seconds, the code will appear in the code container.

Notes

• Each entry in the side menu has an x button to delete it.
• Depending on the DB_TYPE parameter, the side menu items are stored in MongoDB or in a JSON file localted in the db folder.
• You can add additional LLM / Image / Video providers and models in the ./config/app_config.json file, as well as configure all other GSAM parameters.
• All the system prompts used by GSAM are located in the ./config directory.

Screenshots

Main Page

LLM Inference

Suggestions Generation & Model Selection

Suggestion Applied to Prompt

Code Generation

App Ideation Page [1]

App Ideation Page [2]

Presentation Generation

Image Generation

Image Gallery

Video Gallery

Context

This project was developed as part of the Llama Impact Hackathon organized by Lablab.ai.

• Project submission page: GSAM

• Presentation video: Llama Impact Hackathon - GSAM App Video Presentation

• Presentation document: Llama Impact Hackathon - GSAM App Maker Presentation

• Team: The FynBots

Contributors

Carlos J. Ramirez

Please feel free to suggest improvements, report bugs, or make a contribution to the code.

License

This project is licensed under the terms of the MIT license. See the LICENSE file for details.

Acknowledgements

• Meta for developing the Meta Llama powerful models.
• Lablab.ai for organizing the Llama Impact Hackathon.
• Streamlit for providing a user-friendly interface for interacting with the application.
• Open-source community for inspiring and supporting collaborative innovation.
• Users and contributors for their feedback and support.