Composition (Codex's priority — content mix was the louder problem):
- _select_diverse now guards the daily five's emotional tone: at most 1 health,
at most 2 science+health combined, at most 2 of any topic, distinct sources —
so at least three of the five are community/culture/animals/environment when
available. Caps relax (mix, then source) only to fill on thin days.
- Verified live: today's five went to environment x2, health, animals, science.
UI:
- Source moved to its own line below the tags, left-justified, for uniform
rhythm across hero and tiles (was sometimes trailing the tags, right-aligned).
- Watermark kept as-is (intentionally subtle; liked).
Tests updated for the emotional-mix contract (80 total).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Frontend (the premium baseline):
- The hero is now the ONLY image slot. Soft feed images get an atmospheric
gradient overlay; no over-reliance on inconsistent RSS image quality.
- Every secondary/lane card is a uniform typographic editorial tile: no
thumbnails, equal visual weight, a faint topic wordmark watermark, a slim
sage top accent, consistent source, reason text as the trust signal, visible
Replace with quiet tuning actions. Fixes the jarring mixed-media row rhythm
and removes muddy thumbnails entirely.
Backend (composition):
- _select_diverse now balances topics: no more than 2 of one topic while other
topics have candidates (relaxing source then topic caps only to fill), so the
daily five stop clustering medical/science items. Candidates now carry s.topic.
Tests updated for the topic-balance contract (79 total).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- ArticleCard: derive safeHref from article.url and reset image-failure state
when the article changes, so in-place replacements re-evaluate correctly
(clears the Svelte capture warning; build is warning-free again).
- Downweight paywalled stories below readable ones (stable sort) when composing
the daily five and in feed results — the brief now leads readable and rarely
hands over a locked door.
- review_sources gains a 'paywall-heavy' advisory flag (Nature, New Scientist
flag at 100%); never auto-deactivates.
- New Scientist/Nature kept active but no longer reach the daily five; they
remain browsable with the label + Replace.
- Tests: brief readability preference + paywall-heavy flag (79 total).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- paywall.py: conservative domain-level paywall detection (New Scientist,
Nature, and common hard/soft paywalls). Never fetches pages — an honest hint.
- API: Article gains a 'paywalled' flag; the brief now leads with a gentle AND
readable story (paywalled/charged stories stay in the five, just not first).
- New GET /api/replacement returns the next-best readable, unshown article
(honors mood+prefs via the merged prefs param; gentle=true for hero swaps).
- UI: paywalled cards show 'May need a subscription'; a Replace / 'Find one I
can read' action (always visible, while tuning actions stay tucked) swaps the
card for a readable alternative, with a gentle notice when none remain.
- Tests: paywall detection + replacement behavior (77 total).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Today is just the day's five highlights (hero + four) — the preview lanes are
gone; other categories appear only when their mood is selected.
- Each view leads with a large serif heading (Today / Wonder / ...) and a
subtitle, with a quiet sage rule — switching moods retitles the page.
- Drop the now-unused Lane usage from the home.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Text-first lane cards were rendering the full source description, ballooning
them to several times the height of image cards.
- Show the source description only on the hero; lane/grid cards show title + the
one-sentence 'why', which is a tidy uniform summary.
- Line-clamp lane card titles and 'why' to 3 lines (hero description to 6) so
variable text length can't blow up a card. Rows now stay even.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per the calm north star (images support reading, never become a stimulation
layer; metadata-only stays the posture):
- Image-less cards are now designed, not missing: secondary cards are text-first
(no empty media band), and an image-less hero becomes a fully typographic lead
with a faint topic wordmark behind it (CSS attr(data-topic)). No big empty
image space is ever reserved.
- Opportunistic extraction: parse the first <img src> from a feed's
content/description HTML when present, canonicalized — never fetching the
article page. Applies to new ingests (existing rows keep their current image).
- Held by deliberate choice: og:image page enrichment, stock/AI imagery, and any
image-coverage requirement for sources.
Tests: feed HTML image extraction (72 total).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- New BoundariesPanel.svelte: gentle, device-local controls. Avoid words/phrases
first (the trust-critical piece), then 'Paused for now' and 'Always hidden',
each with easy remove. Reassures 'nothing leaves this device'; adding a term
refreshes the brief/feed immediately.
- Quiet 'Boundaries' toggle (active indicator) replaces the old calm bar, keeping
the first viewport calm.
- Wording stays gentle throughout: avoid / pause / hide / boundaries — never
blocked/banned/blacklist.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Hero guardrail (core to the promise, not cosmetic):
- New hero.py: the lead story is chosen with a stricter filter than the rest of
the brief — very low cortisol/ragebait and no grief/medical/violence terms
(cancer, glioblastoma, death, diagnosis, ...). Such constructive-but-charged
stories stay among the five; they just never lead by default.
- /api/brief applies user avoid-terms FIRST, then lead_with_gentle, so personal
boundaries always take precedence over the general guardrail.
- Verified live: the brief no longer leads with a glioblastoma story.
Card polish (per review):
- Secondary cards with no real image are now text-first (no row of empty media
bands); hero still always shows media or a typographic fallback.
- Inline tuning actions are quiet until hover/focus on pointer devices, and stay
visible (softer) on touch — less interface machinery.
Tests: hero safety + lead reordering (70 total).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Show the typographic fallback for missing images too (not only on load
error), driven by component state instead of imperative class mutation —
which also clears the unused-CSS-selector build warning.
- Only render external links for http(s) URLs, else href=#.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- New frontend/ SvelteKit static SPA (Svelte 5), served by FastAPI from
frontend/build (falls back to the legacy page if unbuilt).
- Calm design system: cream/sage palette, serif headlines, generous space,
no urgency colors, gentle motion (respects prefers-reduced-motion).
- Home screen: mood-mode nav (Today/Wonder/People Helping/Solutions/Light
Only/Grounded), the daily brief as a hero + remaining four, browsable mood
lanes, an explicit calm end-state, inline Not today / Less like this / Hide
affordances, and device-local Calm Filters mirroring goodnews/filters.py.
- Backend: moods.py + GET /api/moods (single source of truth for the modes);
FilterPrefs gains max_cortisol/max_ragebait ceilings (for Light Only).
- Push categorical filters (include/mute topics+flavors, ceilings) into SQL in
queries.feed so low-ranked-but-matching items (e.g. discovery for Wonder)
are not truncated by ranking; only avoid-terms stay a Python pass.
- PWA manifest + icon (installable; offline deferred per plan).
- Multi-stage Dockerfile builds the site then serves it from the API.
- Tests: queries.feed categorical filters (63 total). README updated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add source health columns (last_success_at, last_error_at, last_error,
consecutive_failures, review_flag, review_reason) via SCHEMA + migration.
- poll_source maintains them: success resets the failure streak and records the
success time; failure increments it and stores the latest error.
- review_sources() flags active sources that are stale, repeatedly failing,
low-acceptance, duplicate-heavy, or doom-skewed (high cortisol/ragebait) over
a recent window. It is purely advisory: it sets review_flag/review_reason and
never changes the active column (human stays in the loop), clearing the flag
when a source recovers.
- CLI review-sources; cycle runs it as a final step (--no-review to skip);
source-report shows a review line for flagged feeds.
- Tests: healthy/failing/stale/low-acceptance/recovery and never-deactivates.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- New source_candidates staging table (status suggested/quarantined/rejected/
promoted, preview_json snapshot) so untrusted/suggested feeds stay out of the
real ingestion path until reviewed.
- sources.py: save_candidate (re-preview never revives a curator's rejection),
list_candidates, reject_candidate, promote_candidate (copies into sources,
inactive by default — active on approval; never automatic).
- CLI: suggest-source / list-candidates / promote-candidate / reject-candidate.
- API: read-only GET /api/candidates (writes stay CLI-only — no unauthenticated
public write surface yet).
- Fix deprecated ElementTree truth-value test in _parse_rss.
- Tests: candidate lifecycle (save/list/promote/reject, status preservation,
name derivation) — 51 total.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- feeds.preview_feed(): fetch + score a sample WITHOUT persisting; returns
freshness, acceptance rate, cortisol/ragebait/PR averages, and example
accepted/rejected items. With an LLM client it also returns topic/flavor mix
and the model's (accurate) acceptance view.
- CLI 'preview-source URL [--sample] [--classify]'.
- API 'GET /api/source-preview?url=&sample=&classify=' with an http(s)-only
guard (SSRF note left for go-public hardening).
- Site 'Suggest a source' panel with Quick check (heuristic, instant) and Deep
check (model, accurate), rendered DOM-safely.
- Tests: network-free preview_feed tests via monkeypatched fetch (45 total).
- README documents the command, endpoint, and updated roadmap.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add API test layer (TestClient): bad prefs -> 200, mute affects feed,
avoid-term filters, brief filters down, counts match filtered feed.
- Render article cards via the DOM API (textContent) instead of HTML string
interpolation, and only allow http(s) hrefs — defense-in-depth XSS guard for
when the feed faces untrusted sources publicly.
- Refresh the stale README Next Steps to reflect what's done vs ahead.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- API endpoints (feed, brief, category-counts) accept a 'prefs' JSON query
param, parsed tolerantly into FilterPrefs (bad blobs never break the feed).
- Feed over-fetches then applies word-boundary filters in Python and slices to
the page; brief is filtered down (no refill); counts are computed over the
same filtered set so browse numbers match the feed exactly.
- Pause.active() coerces naive datetimes to UTC; FilterPrefs.from_dict skips
malformed pauses and non-string list entries.
- Static site adds the humane ladder (Not today / Less like this / Always hide)
plus a Calm filters panel managing pauses, mutes, and avoid-terms in
localStorage. Nothing leaves the device.
- Tests now 38 (added forgiving-parse and naive-now cases). README documents it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add pytest suite (34 tests) covering scoring thresholds, dedup clustering +
representative selection + time window, brief source/category diversity,
avoid-term phrase matching, and text canonicalization/truncation.
- Rewrite _select_diverse with an explicit, tested contract (best-first, one
per source, backfill, then inject a second category by evicting the
lowest-ranked pick).
- classify_articles now returns attempted/succeeded/skipped (ClassifyReport) so
silent model failures are visible in both the cycle and classify output.
- Fix clean_text truncation to stay within max_len (ellipsis no longer
overshoots).
- New filters.py: canonical FilterPrefs shape (include/mute topics+flavors,
avoid_terms, pauses) and pure word/phrase-boundary matching engine seeding
Calm Filters. Not yet wired into the API.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- cycle now prints per-article classify progress (flushed) so the long step is
clearly alive rather than appearing hung.
- An exclusive flock guards the cycle so a manual run and the systemd timer (or
two timer ticks) cannot overlap and contend on the database and model.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- LocalModelClient.embed() calls the OpenAI-compatible /embeddings endpoint
(local nomic model); base_url shared with chat, model via GOODNEWS_EMBED_MODEL.
- New article_embeddings table and articles.duplicate_of column (+ migration).
- dedup module: embeds missing articles, clusters near-identical stories within
a date window by cosine similarity (pure-stdlib, vectors normalised once), and
marks all but the highest-ranked member of each cluster as a duplicate.
- 'dedup' CLI command; cycle now runs poll -> classify -> dedup -> brief.
- Feed and brief queries hide duplicates, so a story carried by multiple
outlets shows once.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Briefs now fill from a rolling window (prefer today, backfill up to
window_days) and exclude anything featured in the last 7 days of briefs, so
slow days still produce five items without stories lingering day to day.
- New 'check-feeds' command fetches and parses every feed to catch dead ones.
- Added 16 validated sources (science, environment, animals, culture),
expanding coverage from 12 to 28 feeds to reduce staleness.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- deploy/goodnews.service: oneshot unit running 'goodnews cycle' with a
generous TimeoutStartSec so long classify runs are not killed.
- deploy/goodnews.timer: every 15 min, Persistent=true to catch missed runs.
- deploy/goodnews.env.example: LLM endpoint + DB path for the scheduled run.
- README: scheduling/install docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- poll_due_sources(): polls only sources whose last successful poll is older
than their poll_interval_minutes (or never polled), finally giving that
config field meaning.
- classify gains only_unclassified to spend the LLM solely on new (heuristic)
articles, so a frequent scheduled run stays cheap.
- 'cycle' command runs poll-due -> classify-new -> rebuild-today's-brief, with
each step non-fatal so a down model endpoint or empty day never aborts it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- queries.py: shared read-only query helpers (feed, brief, category counts)
returning plain dicts, used by the API and available to the CLI.
- api.py: FastAPI service with Pydantic response models (the companion-app
contract), CORS, and endpoints for categories, feed, brief, and health;
mounts a static site at /.
- static/index.html: minimal dependency-free site rendering the daily five
and topic/flavor category browsing.
- 'goodnews serve' command launches uvicorn (lazy import; core CLI stays
pure-stdlib). Web deps live behind the optional [web] extra.
- Dockerfile + .dockerignore + build-system metadata so the service installs
and deploys cleanly, with the DB mounted as a shared volume.
- README: web/API and deployment docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- New taxonomy module: single source of truth for 6 topics x 5 flavors,
shared by the LLM response schema (enum-constrained) and validation.
- Classifier now assigns one topic + one flavor per article; json_schema
enums force valid values, with coercion as a safety net.
- article_scores gains topic/flavor columns via an idempotent migration.
- New 'list-category' command to browse by topic and/or flavor, ranked by
composite score.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Use json_schema structured output (newer LM Studio rejects json_object),
escalating through json_schema -> json_object -> text and pinning the
first format the server accepts to avoid wasted round-trips.
- Make per-article failures non-fatal and commit incrementally so a single
timeout no longer discards the whole batch.
- Raise default timeout to 180s (configurable via GOODNEWS_LLM_TIMEOUT) for
larger local reasoning models.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
thejayman77