Skip to main content
Ingestion runs a document through the full pipeline: Unstructured.io partitions the text, pdfplumber extracts images, VoyageAI embeds each chunk, and the resulting vectors are upserted into Qdrant.
Ingestion requires a valid Unstructured.io API key and URL configured in your .env file. The pipeline exits with an error if this is not set. See Configuration for details.

​
CLI ingestion

Use the :ingest command inside the running CLI:
:ingest <path> [institution] [course]
1

Run the ingest command

Provide a path to your PDF. Wrap paths that contain spaces in double quotes:
# Simple path
:ingest /home/user/notes.pdf

# Path with spaces
:ingest "/home/user/lecture notes/week3.pdf" MIT biology

# Full metadata
:ingest /home/user/chem101.pdf Stanford chemistry
2

Wait for ingestion to complete

A progress indicator is shown while the document is processed. When ingestion finishes, Quark prints the chunk counts:
✓  notes.pdf — 42 chunks  3 visual
3

Verify with :docs

Run :docs to confirm the document appears in the active session’s ingest log.

​
Arguments

ArgumentRequiredDefaultDescription
pathYes—Absolute or relative path to the PDF. Quote if the path contains spaces.
institutionNo"Default"Tag used to scope vector search. Persists for the rest of the session.
courseNo(none)Free-text label for the course or topic.
Ingest multiple documents tagged with the same institution value, then use that value when querying to scope answers to a specific document set.

Build docs developers (and LLMs) love

Get started for freeTalk to us