* feat(digg): add Digg AI 1000 source module with cluster search and post enrichment
- search_digg shells out to digg-pp-cli with --since 30d --agent
- parse_digg_response normalizes clusters to last30days dict shape
- enrich_with_top_posts attaches top-ranked X posts to top-K clusters
- shutil.which gate plus subproc.run_with_timeout discipline matches
bird_x.py / youtube_yt.py patterns
25 unit tests cover parse, age window, relevance, binary-missing
fallback, timeout recovery, and partial enrichment failures.
* feat(digg): wire Digg source into pipeline, normalize, signals, and render
pipeline.py:
- Import digg, add to MOCK_AVAILABLE_SOURCES, gate via shutil.which
- Dispatch case calls search_digg + parse_digg_response, runs
enrich_with_top_posts at default/deep depth
- Mock fixture includes one enriched cluster + one bare cluster
normalize.py:
- _normalize_digg maps cluster dicts to SourceItem with
container='Digg AI 1000' and metadata.posts pass-through
signals.py:
- SOURCE_QUALITY['digg'] = 0.85 (top tier alongside YouTube,
reflecting Digg's curatorial layer)
- ENGAGEMENT_WEIGHTS['digg'] balances postCount, uniqueAuthors,
and the rank_score derived from Digg's curatorial position
render.py:
- SOURCE_LABELS['digg'] = 'Digg AI 1000'
- _FOOTER_SOURCES adds '⛏️ Digg AI 1000' line after GitHub
- ENGAGEMENT_DISPLAY mirrors footer keys
- New _digg_posts_for + _format_digg_quote helpers emit inline
'@handle via Digg AI 1000' quotes for clusters with attached X
posts; both compact and full-dump renderers call them
* feat(digg): polish per-item engagement display and progress label
- ENGAGEMENT_DISPLAY for digg uses 'posts' / 'auth' to match the
codebase abbreviation convention (HN: 'pts'/'cmt', X: 'rt'/'re')
- Footer item word changes from 'story' to 'cluster' to dodge the
pre-existing naive plural in _footer_line_for_source ('storys')
and to match Digg's actual data model
- ui.py SOURCE_COMPLETION_META adds digg with correct 'cluster'/
'clusters' plural so 'Research complete' shows 'Digg: N clusters'
* feat(digg): document Digg AI 1000 source in skill, README, and changelog
- planner.py SOURCE_CAPABILITIES adds digg with discussion/social/link
capabilities so the planner offers it through the standard fanout
- SKILL.md ACTIVE_SOURCES_LIST gate includes 'which digg-pp-cli' check
and the source list / available-sources line names digg as opt-in
- README.md Sources table adds the Digg AI 1000 row with the activation
gate so first-time readers see what they get
- CHANGELOG.md Unreleased section calls out the source addition
* fix(digg): enrich post-dedupe so brief survivors carry inline quotes
Pipeline dispatch was attaching X posts to the top-3 items returned by
search, but dedupe later picked different survivors when multiple
clusters compared similar (common for trending topics). The brief
ended up showing clusters with no posts attached even though
enrichment ran successfully on positions 0-2.
Move enrichment to _finalize_items_by_source. The new
digg.enrich_source_items helper reads metadata['clusterUrlId'] and
writes metadata['posts'] in place on the SourceItems that actually
survive dedupe.
Verified live on 'openclaw': 2 surviving clusters, both now carry
real X-post quotes from @sama and @jeremyphoward attributed
'via Digg AI 1000'.
Adds 3 unit tests covering survivor enrichment, non-digg skip, and
clusterUrlId fallback to item_id.
* test(digg): relax live off-topic test to check shape, not emptiness
Digg's live search uses fuzzy/popularity fallback, so an impossible
token can still return some loosely-related clusters. The contract
the pipeline depends on is shape (results is always a list);
token-overlap relevance handles the noise downstream.
---------
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Matt Van Horn
c04bd67922