Repository navigation
good-dog demo: corpus expansion (54→62) + reproducible pipeline + LadybugDB graph-sins - #54
Merged
Merged
Conversation
…entary) Group C (Cat 7 token/cost stress, dense docs): AVMA breed-bite literature review, AAFCO nutrient-profile revisions, Ostrander/Parker dog-breed genomics clades, WSAVA nutritional assessment guidelines. Group E (supplementary): FDA xylitol aliases (Cat 4), UK Dangerous Dogs Act + XL Bully (Cat 6), Danish-Swedish Farmdog 202nd AKC breed (Cat 6/5, closes the FSS pipeline), WTTW Chicago shelter overcrowding (Cat 2c/1). Corpus 54 -> 62 notes; +16 ground-truth questions (50 -> 66). validate.py OK, 619 passing. 7/8 substring-verified live (PDFs via pdftotext fallback); E4 (WTTW) authored from manifest-verified facts only — live source is Cloudflare-blocked, flagged for re-verification.
…sins Commits the subagent-swarm graph-build pipeline out of /tmp into the repo so the good-dog demo graph is reproducible and preserved: - tools/good-dog-graph-pipeline/: 8 stage scripts + 5 Workflow scripts (extract / resolve / recover / recall / author-sources) + ontology adapter + the self-contained 62-note viz + a README documenting the flow, deps (LadybugDB 0.17.1, Ollama, Claude Workflow) and the two-layer ground-truth design. - docs/ladybug-graph-sins.md: public-facing LadybugDB operational findings, stripped of machine/project specifics; headlines the 0.17.1 incremental-REL -write corruption repro for the maintainers. - pyproject: exclude the reference build scripts from ruff (terse one-offs, not maintained library code). Pipeline result: 62 notes -> 579 entities / 533 edges, ~87% edge-recall vs ground truth, 0 fabricated edges, largest connected component 289 nodes.
M0nkeyFl0wer
changed the title
- all_rights_reserved (davemech Mech-1970 note) -> fair_use_excerpt: the note short-excerpts a copyrighted source for the contradiction pair (its license_note already states fair-use rationale); matches the convention of the other 28 fair-use-excerpt notes. - cc_by_4_0 -> cc_by_4.0 (formatting dup); fair_use / fair_use_research -> fair_use_excerpt (unify the fair-use family); strip stray quotes/whitespace. - Left distinct licenses intact: public_domain, cc_by_sa, cc_by_nc_4.0, open_government_license_*, and 'other' (proprietary web pages, honest catch-all).
The WTTW Chicago shelter article was confirmed live in a human browser 2026-06-19 (page title matches source_title), establishing the URL resolves to the correct article. The page stays behind Cloudflare for automated fetchers (403, no archive.org snapshot), so exact on-page strings aren't machine-matched; the 265% / 2,455 figures trace to the fact-checked source manifest. Updated the note's provenance paragraph accordingly.
…tion
Opens with the title + one-line tagline + Contents so the substance ('What this
is') comes first; the Fifth Element image and the Multipass/MemPalace origin
blockquote move to the bottom.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Shareable good-dog graph-demo package.
1. Corpus expansion — Group C + E (54 → 62 notes)
Completes the expansion manifest (A/B/D already in v0.3).
2. Reproducible graph pipeline (
tools/good-dog-graph-pipeline/)The subagent-swarm build harness, committed out of /tmp: 8 stage scripts + 5 Workflow scripts (extract → resolve → recover → recall → author-sources) + ontology adapter + the self-contained 62-note viz + a README. Extraction is a Claude subagent swarm (not a local model); embeddings stay local; writes use LadybugDB reconstruct-and-swap.
3. Public LadybugDB graph-sins (
docs/ladybug-graph-sins.md)Operational corruption/crash findings + safe primitives, stripped of machine/project specifics. Headlines the 0.17.1 incremental-REL-write corruption repro for the LadybugDB maintainers.
Result
62 notes → 579 entities / 533 edges, 86% edge-recall (405/469) vs ground truth, 0 fabricated edges, largest connected component 289 nodes.