Purpose

Why Inpedia exists — local, private, and controlled.

Local-first personal knowledge system. Ask questions; get grounded answers with citations. Corpus and embeddings stay on personal hardware only.

Motivation

Why build this

ChatGPT

Upload notes and ask questions in a hosted chat product.

Subscription cost, usage limits, corpus on a vendor platform.

Confluence

Pay Atlassian to host a personal or team wiki.

Per-seat SaaS rent — storage and search, not local RAG.

Inpedia

Local-first index on your own hardware.

You operate it — but privacy, control, and zero marginal query cost.

What it is

Inpedia is a personal professional operating system — a conversational index over my own files, documents, and notebooks.

It is not a hosted service. This repository and public site describe architecture and code, not corpus content.


Pipeline

How it works (today)

  1. 1

    Capture

    Personal notes as dated Markdown on your Mac.

  2. 2

    Ingest

    Chunk → embed via Ollama → store in Chroma.

  3. 3

    Ask

    Retrieve excerpts → grounded answer with citations.

Boundary

Public vs local

Published (site, git)Local only (`Documents/Inpedia Knowledge Base/`)
Code, ADRs, changelog`Source notes/` — Markdown notebooks
Architecture docs`Search index/` — Chroma vector store
`Inbox/` — Review draft exports
`Originals/` — immutable scan archive
`review.py`, `ingest.py`, `query.py`, `chat.py`

v1

Stack

OllamaChromaPythonApple Vision OCRMarkdown + photographed notebook pages

Per-component rationale in architecture decisions.