Skip to content

dotpredicate/myAI

Repository files navigation

myAI

Like Open WebUI but better suited for me.

Project description

Goals

  • integrate llama.cpp for running local models
  • create a great RAG / DB / search engine for AI
  • simple and powerful agent workflow creation
  • add permissions, security policy and a safety model

Guidelines

  • keep it simple
  • just works (tm)
  • secure and private
  • just use postgres

Technical info

Project structure:

  • ./app.py - FastAPI routes and application entrypoint
  • ./conversation.py - conversation state management and orchestration
  • ./inference.py - AI protocol handling (stateless)
  • ./index.py - search indexing and vector search logic
  • ./database.py - database connection and utility functions
  • ./documents.py - document processing and management
  • ./system.py - system-level utilities
  • ./tools.py - tool execution logic
  • ./migrations/ - folder with migration scripts
  • ./static/ - static resources deployed by the server, web front-end

Quickstart

  1. source .venv/bin/activate
  2. pip install -r requirements.txt
  3. python app.py

Server runs at http://127.0.0.1:5000

Migrations

At startup, the module scans the migrations folder for SQL files, sorts them in lexicographic order, executes those that have not yet been applied and records the execution in the migrations table

Code style

  • use Python types

Simplifications for now

  • web UI only, no mobile, minimal styling
  • single user, no log-in
  • no cloud integrations
  • monolithic back-end, single instance
  • no caching like Redis, prefer in-memory structures

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors