AI Diary

Idea

Parse diary in PDF files, add tags, keywords, build analytics using LLM.

Architecture

• use Langchain as framework to connect LLM with python
• use Gemini as LLM
• use Postgres as Vector DB and DocStore

Getting started

• Install Docker and Docker Compose
• clone project
• run make setup

Everyday Usage

• make start
• make stop
• make api-console - open console of api container
• make invoke-llm QUERY="<QUESTION>" - ask LLM a question. Example: make invoke-llm QUERY="how are you?"
• make alembic-revision M="<MESSAGE>" - generate alembic revision
• make import-diary-file FILE="<FILE_NAME>" - import markdown file from /diary_files
• make import-sleep-csv FILE="<FILE_NAME>" - import csv file from /health_files
• make get-llm-responses - process all records without llm response
• make process-llm-responses - process all records with llm response and generate tags, subjects, locations, etc.

Help

• run make help to see all commands

Endpoints

• http://localhost:8080/ - Adminer database manager
• http://localhost:8000/api - Api
• http://localhost:8000/docs - Api docs
• http://localhost/ - Front

License

TODO

TODO

• AI agent decide if generate chart for data.
• AI agent to decide if it should ask vector db or text to sql.
• Vector DB to enable seaching in diary
• add sports data
• import huawei data