DO NOT TRUST THIS README. I'LL FIX IT. COME TO DuckTapeDevOps

NeuroQuack

NeuroQuack is an innovative platform combining the power of conversational AI and advanced image synthesis to create a unique interactive experience. It allows users to upload images and engage with an AI to transform those images into personalized caricatures. Dive into a generative world where every conversation with the AI leads to a new visual discovery.

MVP: Minimal Viable Product

Docker Container running locally for Twitch Bot with interaction with chat
Deployable to AWS
Walkthrough

FMVP: Finals MVP

Integration with a Large Language Model (LLM) for dynamic conversation flow.
Utilization of Stable Diffusion for real-time image generation based on LLM prompts.
AWS SageMaker for hosting and auto-scaling the LLM and Stable Diffusion models.

Tech Stack (Needs)

The NeuroQuack tech stack is meticulously curated to provide a robust, scalable, and responsive user experience:

Frontend: Cybernetic-themed UI, empowering users to craft their own visual journey.
AI: Integration with a Large Language Model for engaging conversation and Stable Diffusion for dynamic image generation.
Cloud Infrastructure: AWS SageMaker for deploying models, and Amazon S3 for data storage, all seamlessly orchestrated using Massdriver to streamline our infrastructure as code, ensuring quick, reliable, and repeatable deployments.
Networking: Amazon Route53 for DNS management, providing a smooth and accessible user experience.
Security and Compliance: Leveraging AWS IAM for robust access control, safeguarding user data and interactions.

Brain Dump (Wants)

Future enhancements we're considering:

Enhanced UI/UX for the IPython Notebook to make it even more immersive.
Expanded LLM capabilities for richer conversations and more accurate prompt suggestions.
Advanced image editing features post Stable Diffusion processing.
Integration with additional AWS services for monitoring, logging, and automated deployment pipelines.

Walkthrough

Your bot Access Token can be found at https://twitchtokengenerator.com

Run make docker_build
Run make docker_run
To start the bot, use the following cURL:

curl --location 'http://localhost:4000/start_bot' \
--header 'Content-Type: application/json' \
--data '{"twitch_token": "<<YOUR_BOT_ACCESS_TOKEN>>", "initial_channels": "<<YOUR_CHANNEL_NAME>>"}'

In your Twitch chat, send !hello and you should receive Hello {ctx.author.name}! from your bot in the Twitch chat!
To stop the bot, use the following cURL:

curl --location --request POST 'http://localhost:4000/stop_bot' \
--header 'Content-Type: application/json'

Run docker logs <<container_id>> to see any logs from the running container locally
Run docker stop <<container_id>>

Backlog

Here's what we're tackling next:

Full mobile responsiveness for the IPython Notebook interface.
Improved error handling and user feedback for image uploads and generation.
Incorporation of user accounts and session management to save and retrieve past interactions.
Enhanced analytics to understand user engagement and model performance.

Working Examples:

- SDXL
  - ECR: 763104351884.dkr.ecr.us-east-1.amazonaws.com/stabilityai-pytorch-inference:2.0.1-sgm0.1.0-gpu-py310-cu118-ubuntu20.04-sagemaker
  - S3: s3://jumpstart-cache-prod-us-east-1/stabilityai-infer/prepack/v1.0.1/infer-prepack-model-imagegeneration-stabilityai-stable-diffusion-xl-base-1-0.tar.gz
  - Instance Type: ml.g5.4xlarge

Local Testing Commands:

uvicorn main:app --reload to run FastAPI locally and reload with changes for testing

DocLock17/NeuroQuack