ObsidianDataWeave AGENTS.md

ObsidianDataWeave Agent Contract

Purpose

This repository converts source .docx documents and existing Obsidian notes into Zettelkasten-style notes with MOC structure.

Agents should treat this file as the canonical integration contract for both Claude Code and Codex.

Supported Workflows

1. Process a source document into atomic notes and MOC: python3 scripts/process.py "Document.docx" Safe automation form: python3 scripts/process.py "Document.docx" --non-interactive --on-conflict skip
2. Process a curated NotebookLM notebook into atomic notes and MOC: python3 scripts/process_notebook.py "<notebook_id>" Safe automation form: python3 scripts/process_notebook.py "<notebook_id>" --non-interactive --on-conflict skip Optional inputs: --include-sources, --include-mindmap, --profile <name>.
3. Process an existing personal note in the vault: python3 scripts/process_note.py "Note Title" Safe automation form: python3 scripts/process_note.py "Note Title" --mode atomize --non-interactive --on-conflict skip
4. Process a contacts/networking note into individual contact cards: python3 scripts/process_contacts.py "Contacts Note" Safe automation form: python3 scripts/process_contacts.py "Contacts Note" --non-interactive --on-conflict skip
5. Generate markdown files from an existing atom plan JSON: python3 scripts/generate_notes.py /path/to/atom-plan.json
6. Copy staged markdown files into the vault: python3 scripts/vault_writer.py --staging /path/to/staging --atom-plan /path/to/atom-plan.json
7. Find duplicate note candidates or run semantic dedup: python3 scripts/dedup_vault.py --dry-run
8. Run environment checks before operating on the vault: python3 scripts/doctor.py
9. Run deep research directly into a NotebookLM notebook (bypasses the upstream CLI retry duplication bug, see "Why research_notebook.py exists" below): python3 scripts/research_notebook.py run "<notebook_id>" "<query>" Safe automation form: python3 scripts/research_notebook.py run "<notebook_id>" "<query>" --non-interactive Dry-run preview of what would be imported: python3 scripts/research_notebook.py run "<notebook_id>" "<query>" --dry-run
10. Clean up duplicate/error sources in an existing NotebookLM notebook: python3 scripts/research_notebook.py dedupe "<notebook_id>" --dry-run Safe automation form: python3 scripts/research_notebook.py dedupe "<notebook_id>" --include-error --non-interactive
11. Initialize a new LLM Wiki space: python3 scripts/wiki_init.py <slug> --mode project --title "Project" Modes: project (fixed core pages) or corpus (entities-only). Add --lang ru (or --lang en) to pick template language; defaults to [wiki].default_lang in config.toml (en if unset).
12. Ingest raw inputs into a wiki-space: python3 scripts/wiki_ingest.py <slug> <file-or-dir> --kind {articles|docs|transcripts|assets}
13. Compile a wiki-space (LLM merges raw into pages): python3 scripts/wiki_compile.py <slug> --since-last-compile Safe automation form: python3 scripts/wiki_compile.py <slug> --since-last-compile --on-conflict overwrite
14. Update one page from a single new raw input (incremental merge): python3 scripts/wiki_update.py <slug> raw/docs/<file>.md
15. Lint wiki-space integrity: python3 scripts/wiki_lint.py [<slug>] [--strict]
16. Search the FTS5 vault memory (lexical full-text over every note): python3 scripts/memory_index.py search "<query>" --json Useful flags: --limit N, --prefix (last term as prefix), --folder X, --tag Y, --raw (raw FTS5 syntax). The index updates automatically after each vault_writer write ([memory].auto_update).
17. Build or refresh the memory index / run the upgrade migration: python3 scripts/memory_index.py build | update | status python3 scripts/migrate.py (idempotent: adds [memory] config, builds index)