Corpus & local home

Personal notebooks live in Documents/Inpedia Knowledge Base — outside this repository.

Personal notebooks and the search index live outside the public Inpedia repository — in a normal folder you can see in Finder.

Two roots of trust

WhatWhere
Code, docs, ADRs, BlueprintGit repo (inpedia)
Markdown source notesDocuments/Inpedia Knowledge Base/Source notes/
Chroma search index (rebuildable)Documents/Inpedia Knowledge Base/Search index/
Review draft exportsDocuments/Inpedia Knowledge Base/Inbox/
Immutable scan originalsDocuments/Inpedia Knowledge Base/Originals/
OCR correction crops (training)Documents/Inpedia Knowledge Base/Corrections/

Decision record: ADR-008

Gitignore was never the wall — it was a seatbelt. The wall is structural: the public repo has no corpus directory, ingest refuses repo paths, and CI enforces the boundary.

Finder: open Documents → Inpedia Knowledge Base.


First-time setup

cd inpedia
python3 init.py
python3 ingest.py tests/fixtures/sample-corpus.md   # fictional pipeline test
python3 query.py "Why was scope frozen before expansion?"

init.py creates the Knowledge Base folder and subfolders in Documents.


Your notebooks

  1. Put Markdown in Documents → Inpedia Knowledge Base → Source notes (or drag files there in Finder).

Use dated section headers:

## 2026-06-20

Your text here…
  1. Ingest from the inpedia repo:
python3 ingest.py your-notebook.md

Bare filenames resolve against Source notes.

  1. Query or chat:
python3 query.py "your question"
python3 chat.py

Migrating old locations

If you used data/raw/ in the repo or the earlier hidden ~/.inpedia/ folder:

python3 scripts/migrate_corpus.py --dry-run   # preview
python3 scripts/migrate_corpus.py

Or drag files manually into Source notes in Finder, then run python3 ingest.py your-notebook.md.


Environment overrides

VariableDefault
INPEDIA_HOME~/Documents/Inpedia Knowledge Base
INPEDIA_CORPUS_ROOT$INPEDIA_HOME/Source notes
INPEDIA_INDEX_ROOT$INPEDIA_HOME/Search index
INPEDIA_INBOX_ROOT$INPEDIA_HOME/Inbox

Use only if you want a different disk or synced folder. Source notes still never belong in the public repo.


What may live in git

  • Fictional fixturestests/fixtures/sample-corpus.md for pipeline and CI
  • Architecture — docs, ADRs, Blueprint
  • Never — personal dated notebooks or Chroma data

Backup

  • Source notes — Time Machine, iCloud, or any backup of Documents/Inpedia Knowledge Base/Source notes/
  • Search index — optional; delete and re-ingest from source notes if lost

Markdown is canonical. The search index is expendable.