Uncle - AI Playground

Goals:

• Generate or edit images using AI models.
• Chat-like interface like ChatGPT for conversational interactions with the AI.
• Allow choosing AI providers and models. (OpenAI and Gemini for now)

Tech Stack

• Axum and Askama for the web server and templating.
• Yaas API for single sign-on and user management.
• OpenAI API for AI interactions.
• AWS S3 for image input/output storage.
• Bulma CSS for multi-page application styling of the website.
• React and Tailwind CSS for the AI Playground interface.

Models

image_prompts:

• id
• user_id
• prompt
• short_title
• model
• background
• moderation
• qty
• output_compression
• output_format
• quality
• status (pending|completed|failed)
• created_at
• updated_at

images:

• id
• user_id
• prompt_id
• category (input|output)
• filename
• file_type
• file_size
• file_path
• dimensions
• created_at

jobs:

• id
• job_type
• prompt_id
• status
• created_at
• updated_at

Image Workflow

• User submits a prompt through the web interface along with attached images if there are any.
• Images are uploaded to AWS S3 using a presigned URL.
• Prompt is received in the server, stored in the database with pending status, sent to a job and returns immediately.
• Background worker sends the prompt to the AI provider and waits for the response.
• Depending on the response, the job is either marked completed or failed.
• Clients are notified with the status change via Websockets.

Job Queue

• On submit of a prompt, a job is created.
• A worker process picks up pending jobs and processes then sequentially.
• Once job is completed or failed, prompt status is updated.

DB Migrations

tursodb db/uncle.db < migrations/migration-file.sql

API

API endpoints:

• GET /api/image-prompts - List all image prompts for the current user.
• POST /api/image-prompts - Create a new image prompt.
• GET /api/image-prompts/{id} - Get details of a specific image prompt.
• GET /api/image-prompts/{id}/status - Get the status of a specific image prompt.
• POST /api/images/upload-urls - Get presigned URLs for uploading images to S3.
• POST /api/image-prompts/{id}/images - Add a new image related to the image prompt.
• GET /api/image-prompts/{id}/images - Get all images associated with a specific image prompt.

Response envelope:

• Success: { status_code: u16, message: String }
• Error: { status_code: u16, message: String }