Skip to content

docs(readme): mission, honest extraction numbers, NARA scans, no em-dashes - #106

Merged
JordanCoin merged 3 commits into
mainfrom
docs/readme-honest
Oct 19, 2026
Merged

JordanCoin merged 3 commits into
mainfrom
docs/readme-honest

Conversation

@JordanCoin

Copy link
Copy Markdown
Owner

Four README fixes:

  1. Mission under the title. The "anyone, anywhere, investigate power safely" paragraph from CLAUDE.md, so a reader knows why local-first and purgeable matter before the feature table.
  2. Honest extraction numbers. "99.5% accuracy across 480K chars in 50 PDFs" and "100% on every FOIA document tested" had no reproducible source (glyph-api fix: load cached GLiNER model offline before falling back to download #81). Replaced with what the engine's quality ledger records: 46 PDFs, 2.9M characters, 0.0005% unresolved glyphs, 45/46 clean flow, 38/46 no detectable corruption. "100% recall with a 2GB local model" (from a one-document synthetic benchmark) replaced with the tier fallback description.
  3. NARA row now says the scans themselves download and OCR locally (--objects) and a series can be walked (--ancestor --all), per feat(nara): child-record walks and digital-object downloads, like NARA's scripts #104.
  4. No em-dashes or arrows (nine and two removed).

tests/test_plugin.py still passes (README commands unchanged).

🤖 Generated with Claude Code

https://claude.ai/code/session_01Wdqj51cPquyTRKmoJLQV6y

…no em-dashes

Replaces the unsourced '99.5% accuracy across 480K chars' and '100% recall'
claims with the figures the engine's quality ledger actually records
(glyph-api #81), states the mission from CLAUDE.md under the title,
updates the NARA row for --objects and --ancestor, and removes the nine
em-dashes and two arrows.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wdqj51cPquyTRKmoJLQV6y
Copilot AI balanced review requested due to automatic review settings October 19, 2026 09:31

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

JordanCoin and others added 2 commits October 19, 2026 20:26
The 0.7.0 ledger figures came from a non-deterministic engine on a
46-file corpus. Replace them with the 0.8.0 sweep: 176 PDFs, 15,808
pages, 85 percent clean text layer, the rest flagged for OCR.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wdqj51cPquyTRKmoJLQV6y
@JordanCoin
JordanCoin merged commit f6e5ef7 into main Oct 19, 2026
7 checks passed
@JordanCoin
JordanCoin deleted the docs/readme-honest branch October 19, 2026 18:44
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants