Llama Impact Hackathon

Start

The FynBots

Carlos J. Ramirez

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

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.


"The FynBots team on Llama Impact Hackathon Hackathon"

Team Idea

Submission