Ingest Files¶

Lattice can extract atoms from documents — PDFs, Word docs, spreadsheets, markdown, and plain text.

Supported file types¶

Extension	Notes
`.pdf`	Text extracted via `pypdf` — requires `uv sync --group pdf`
`.docx`	Requires `python-docx`
`.pptx`	Text from slides
`.xlsx`, `.xls`	Cell content as text
`.md`	Segmented by headings
`.txt`	Plain text

Method 1: `lc` CLI¶

uv run lc path/to/meeting-notes.md
uv run lc ~/Downloads/report.pdf

Prints atom count on completion.

Method 2: Web UI drag-and-drop¶

Open localhost:7337 and drag a file onto the chat area. The web UI uploads it to /api/ingest-file and shows the result.

Method 3: Inbox file drop¶

Copy any .md or .txt file into LATTICE_DIR/inbox/:

cp ~/notes/book-summary.md ~/.lattice/inbox/

The daemon picks it up within seconds and moves it to processed/. This is the best path for automation or bulk ingestion.

Note

Only .md and .txt files are supported by the inbox watcher. For PDF/docx, use lc or the web UI.

Install PDF support¶

uv sync --group pdf

Without this, .pdf files will fail with a missing dependency error.

Source ID¶

For file ingests, the source_id field on each atom is set to the filename (e.g., meeting-notes.md). This makes it easy to filter atoms by source in the graph.