Revolutionary War
Artifact Finder
Find historical artifacts hiding in centuries-old text.
Historians and collectors spend months manually reading diaries, journals and newspapers to locate artifact mentions. This tool ingests OCR'd primary sources, uses an AI extraction pipeline to pull out named, location-bound artifacts with the exact supporting quote, geo-resolves them, and scores every finding by confidence.
$ pe listfindings --min-conf 0.70
● musket · Elizabethtown 1776 · conf 0.86
"…a fine musket taken from the redcoats…" — Fithian Journal
● regimental flag · Trenton · conf 0.78
● supply cache · Monmouth · conf 0.72
✔ 386 findings extracted · Cumberland Co. corpus
Representative output — not a live instance.
●Ingests primary sources — pulls OCR'd historical texts (journals, archives, newspapers) and prepares them for analysis.
●AI artifact extraction — finds named, location-bound physical artifacts (weapons, documents, flags, caches) and captures the verbatim evidentiary quote.
●Geo-resolution — maps mentioned places to coordinates via a built-in gazetteer of Revolutionary War sites.
●Confidence scoring & dedup — clusters mentions into single findings scored 0–1 by name similarity and location.
●Watchlists & alerts — define a geographic area of interest and get emailed on new high-confidence discoveries; browse everything on an interactive map.
0
mapped RevWar sites
0
counties in gazetteer
0+
source pages ingested
map + alerts
live findings dashboard
Python · FlaskSQLiteReact 18AI extraction · confidence scoringemail alerts
ℹ
Not offline: extraction uses a hosted AI API (Anthropic Claude) and ingestion pulls from public archives — it needs an API key and a network connection to run. Confidence scores are the model's own, heuristically adjusted, not validated against ground truth. Currently a single-source (Fithian Journal) prototype.
🔒
Capabilities and sample findings shown. Source, extraction logic and data are private — this is a capability showcase, not a distribution.
Buried insight
in a pile of documents?
I build tools that read what humans don't have time to — and surface the findings with evidence attached. Tell me your corpus.
Start a project →