LORE

Knowledge base management stack. Using:

• Local LLM
• Obsidian
• Raycast

The goal is to augment a local LLM with my own Obsidian notes. If I ask a question that is not yet covered in my notes, offer the ability to generate a note. This should be available quickly and without friction, hence the decision to use the Spotlight replacer, Raycast.

implementation

graph LR
    Q["User asks question"] -->r
    r["retrieve notes"] --> gd["grade notes"]
    r -.- db
    db[(Embeddings vector storage)]
    gd --> a["answer question"]
    a -->rn{Where there relevant notes?}
    rn -.->|True|e[Show answer]
    rn -.->|False|ui{does user want to generate note?}
    ui-.->|True|gn["generate note"]
    ui-.->|False|e
    gn -->e

Loading

Note format

I use a format I have taken to calling atomic notes.

An example may be:

what is postgres.md
---
tags: ['postgres', 'database']
---

PostgreSQL, often abbreviated as Postgres, is a powerful open-source
relational database management system (RDBMS) that allows you to store,
manipulate, and retrieve data efficiently. It's a popular choice for
building scalable and reliable databases, especially in the enterprise
environment.

TODO

• Link to notes used
• Actually create Obsidian note
• Implement Raycast extension
• Implement structured output