Literature monitoring

Citracer can compare a new trace against a previous one to highlight what changed, turning a one-shot snapshot into a monitoring workflow.

Usage

# Step 1: initial trace with JSON export
citracer --pdf paper.pdf --keyword "attention" --depth 3 --export baseline.json

# Step 2: months later, re-run and diff
citracer --pdf paper.pdf --keyword "attention" --depth 3 --diff baseline.json

New nodes are colored orange and labeled NEW in the info panel. The legend gains a clickable "new (since last run)" row to toggle them.

Date filtering

--since highlights nodes by publication date, with or without --diff:

# Papers published 2025 or later (standalone)
citracer --pdf paper.pdf --keyword "attention" --since 2025

# Papers published June 2025+ AND not in the baseline (intersection)
citracer --pdf paper.pdf --keyword "attention" --diff baseline.json --since 2025-06

Flags	A node is "new" when
--diff only	Not in baseline
--since only	Published on or after the date
Both	Both conditions met

Month-level precision uses the publicationDate field from Semantic Scholar (YYYY-MM-DD). When only the year is known, --since 2024-06 falls back to year-only comparison.

Note

Nodes with no known publication date are skipped by --since and a warning is logged with the count.

Export

Both is_new flags (on nodes and edges) are included in JSON and GraphML exports, so downstream scripts can consume the diff without re-running citracer.

Known limitation

paper_id is not fully stable across runs. If a paper was resolved by title hash in one run and by DOI in another, it may falsely appear as "new". Re-running both traces from the same cache directory minimizes this.