Skip to content

Guide Quality & Category Discovery — high-confidence classification, category hubs, and couch-first What's On views #191

Open

Description

@Jumpstile

Parent

Product-owner priority

Guide quality is the most important product-quality dimension.

A beautiful interface over incomplete, stale, misclassified, duplicated, or misleading programme data is a failed product.

ChannelForge must optimize first for:

  • correctness
  • completeness
  • freshness
  • classification quality
  • event/program identity
  • deduplication
  • clear provenance/confidence
  • fast discovery from the couch

The presentation layer should then turn that high-quality guide into a set-top-box experience that is easier to browse than a conventional cable grid.

Product outcome

The ChannelForge living-room interface should provide category-first discovery surfaces such as:

  • Wrestling
  • Football
  • Baseball
  • Basketball
  • Hockey
  • Soccer
  • Motorsports
  • Combat Sports
  • Other Sports
  • Movies
  • TV Shows / Series
  • News
  • Kids / Family
  • Music
  • Live Events
  • Local
  • Favorites / My Stuff

The initial category set may evolve based on real usage, but sports and wrestling must be first-class categories rather than buried under a generic Sports bucket.

Home / category-hub concept

The home surface may use large remote-friendly tiles, for example:

  • Guide
  • Live Now
  • Wrestling
  • Football
  • Baseball
  • Soccer
  • Movies
  • News
  • Search
  • Apps

Selecting a category opens an optimized discovery surface rather than forcing every category into the same UI.

Examples:

Wrestling

  • On Now
  • Tonight
  • Coming Up
  • WWE
  • AEW
  • TNA
  • NJPW
  • MLW / other promotions
  • Recently available / catch-up entry points where supported

Football

  • Live Now
  • Starting Soon
  • NFL
  • College
  • Other leagues
  • Team-focused filtering/favorites

Movies

  • On Now
  • Starting Soon
  • New / premieres where supported
  • By provider
  • By genre
  • Search

News

  • Live Now
  • Local
  • National
  • Business
  • International
  • Weather / specialty where useful

These are examples, not rigid final layouts.

Presentation research / invention requirement

Do not assume the conventional horizontal channel-grid is the final answer.

Evaluate and prototype couch-first presentations such as:

  • large category tiles
  • Now / Next cards
  • vertical event lists
  • timeline rails
  • compact EPG grids
  • poster/artwork rows where appropriate
  • league/promotion/team grouping
  • provider-neutral event cards
  • “starting in X minutes” sections
  • daypart sections
  • hybrid layouts

Different categories may legitimately use different presentation patterns.

The chosen interaction must prioritize:

  • Harmony D-pad efficiency
  • minimal clicks
  • strong focus visibility
  • readable distance typography
  • quick jump to Now
  • quick return Home
  • obvious “Watch” action
  • simple alternate-source access
  • no dependency on pointer/mouse interaction

Guide classification model

Guide content must be categorized from the canonical model, not UI-only string matching.

Classification should support, where applicable:

  • content type
  • sport
  • league
  • competition
  • team
  • promotion
  • event type
  • movie
  • series/episode
  • news
  • kids/family
  • music
  • live/scheduled status
  • provider
  • channel/network
  • region
  • language
  • user tags/favorites

Each classification should preserve evidence/provenance and confidence where inferred.

Guide quality gates

At minimum measure/report:

Accuracy

  • correct title
  • correct start/stop time
  • correct channel/provider association
  • correct event identity
  • correct teams/promotion/league/category

Completeness

  • expected channels/events represented
  • missing schedule coverage detected
  • missing metadata identified explicitly

Freshness

  • fetch time
  • source age
  • stale-window policy
  • changed/postponed/cancelled events updated promptly

Deduplication

  • one canonical event/program when multiple providers carry it
  • alternate offerings retained without duplicate user-facing event cards

Classification quality

  • category and subcategory confidence
  • no silent cross-category contamination
  • ambiguous classification routed to review/unknown rather than guessed

Availability quality

  • distinguish “scheduled” from “actually available to this user”
  • retain provider/source evidence
  • do not claim entitlement where unknown

Quality-first source policy

Provider-supplied EPG is evidence, not automatic truth.

ChannelForge should use the evidence model from #40/#175 and event intelligence from #28/#30 to:

  • compare overlapping sources
  • prefer stronger/fresher evidence
  • detect disagreement
  • repair only when sufficiently supported
  • preserve original values and correction provenance
  • fail visibly rather than fabricate

User-visible trust cues

The normal user should not see technical confidence scores everywhere, but the UI should make meaningful uncertainty understandable:

  • Confirmed
  • Updated
  • Source unavailable
  • Schedule may be stale
  • Needs attention

Advanced details can expose source/provenance/confidence.

Performance requirement

Category entry should feel immediate on the SER8.

Precompute/index the guide as necessary so normal operations such as:

  • open Wrestling
  • open Football
  • show Live Now
  • show Tonight
  • move between events
  • open details

do not trigger slow provider fetches in the foreground.

Refresh and normalization belong in the background through #35.

Acceptance criteria

  • Category taxonomy is represented in the canonical model, not hard-coded solely in UI.
  • At least Wrestling, Football, Baseball, Soccer, Movies, and News have dedicated couch-first views.
  • Wrestling has promotion-aware grouping and high-quality schedule/event identity.
  • Sports categories are league/team-aware where data permits.
  • Each category supports On Now and Coming Up.
  • At least two presentation patterns are prototyped/tested before locking the final 10-foot UX.
  • Conventional EPG grid remains available where it is useful, but is not the only discovery model.
  • Cross-provider duplicate events appear once with multiple offerings.
  • Guide-quality metrics and failure states are testable.
  • Stale/incomplete/ambiguous guide data is surfaced honestly.
  • Harmony Elite navigation is complete without routine keyboard/mouse use.
  • Category views integrate with Universal Live Guide — one normalized timeline across linear TV, live streaming, sports, and scheduled events #186 canonical guide and SER8 Living-Room Reference Client — appliance UX, Harmony control, and platform selection #188 SER8 client rather than creating separate guide stores.
  • Exact-head docs/screenshots reflect the shipped interaction.

Release impact

This issue is release-driving under #190.

Provider breadth must not outrank guide quality. If schedule pressure forces a tradeoff, prioritize a smaller number of well-supported providers with excellent guide accuracy/classification over a larger provider list with weak or misleading guide data.

Activity

  1. added
    priority:highShould be next; significant impact
    component:uxGUI, review workflows, previews, and user intake
    on Sep 27, 2026
  2. Jumpstile commented on Sep 27, 2026

    @Jumpstile
    OwnerAuthor

    Product-owner interaction requirement: category hubs must be directly actionable, not just informational.

    Example — Wrestling:

    1. Open Wrestling.
    2. See a provider-neutral list of what is Live Now / Starting Soon / Tonight / Coming Up, including shows, events, PPVs, PLEs, specials, movies/documentaries where classified as wrestling content.
    3. Highlight an item and press OK/Watch.
    4. ChannelForge resolves the best currently available authorized offering for that canonical item.
    5. The SER8 client opens the correct provider/channel/app/browser target directly.
    6. If more than one valid source exists, the default action uses the preferred source; Info/Options exposes alternates.
    7. If no source is currently playable, show the actual reason (not entitled, starts later, blackout/region, provider unavailable, launch unsupported, unknown) rather than failing silently.

    The same interaction contract should apply to Football, Baseball, Soccer, Movies, News, and other category hubs.

    Category views therefore need actionable canonical item IDs plus resolved availability/launch metadata from #70/#67, not presentation-only programme cards.

  3. Jumpstile commented on Sep 27, 2026

    @Jumpstile
    OwnerAuthor

    Guide-quality refinement: category hubs must be both trustworthy and instant-feeling. #193 adds concrete latency budgets for opening Wrestling/Football/Movies/News, Live Now, details, search, and source resolution on the SER8. Guide refresh/enrichment must not block ordinary browsing.

  4. Jumpstile commented on Sep 27, 2026

    @Jumpstile
    OwnerAuthor

    Category/guide UX must satisfy #194: category hubs should be understandable at a glance without instruction, visually polished, readable at TV distance, and tested with untrained users. Guide density must not regress into a desktop-style EPG merely because more metadata is available.

  5. Jumpstile commented on Sep 30, 2026

    @Jumpstile
    OwnerAuthor

    Owner scope update — 2026-09-30

    The expanded One Guide built-in taxonomy is approved and should move forward: live-now, starting-soon, football, baseball, basketball, hockey, soccer, wrestling, motorsports, boxing, mma, tennis, golf, rugby, cricket, lacrosse, other-sports, movies, news, kids, entertainment, documentary, comedy.

    Generic Sports should map to other-sports only when no more-specific recognized sport category is present; unknown XMLTV categories remain omitted.

    PR #199 has merged into main at f20c7f034bb473ef196cddeb9ffaa2399675c227. That merge contains the base One Guide API and unavailable-vs-valid-empty fix, but not this expanded taxonomy. The preserved expansion work should continue as a focused follow-up under #191 while #200 proceeds in parallel in its own worktree.

  6. Jumpstile commented on Sep 30, 2026

    @Jumpstile
    OwnerAuthor

    Coordination note — active parallel lanes

    All lanes share C:\Temp\CHANNELFORGE-AUTHORITATIVE-STATE.txt as the current owner coordination state.

  7. Jumpstile commented on Oct 2, 2026

    @Jumpstile
    OwnerAuthor

    Official wrestling source research — bounded adapter plan (2026-10-02)

    This fills the source-specific gap under #191 and complements the existing canonical evidence/provider/matching work in #29, #30, and #175. It is research and a proposed sequence only: no new epic, scope rewrite, implementation, or release certification. Wrestling remains required across WWE, AEW, TNA, NJPW, ROH, and MLW; phasing AEW first and WWE next does not make the other promotions optional.

    Promotion-owned sources reviewed

    • AEW — first bounded adapter: events calendar, canonical event-detail example, and dated card/preview example. Calendar/detail pages expose event title, date, city, venue and a displayed time; preview/news articles enumerate announced match participants/stipulations and may distinguish venue time from the live broadcast start. The PPV guide gives provider/territory examples per event. Use canonical event URL as the initial dedupe key; no public API ID was established.
    • WWE — next adapter: event finder, official upcoming-event schedule, dated PLE card/preview articles (example Money in the Bank card page), and the where-to-watch page. Event rows/releases provide dates, cities, venues and often local start times; card articles carry matchups/stipulations; watch information varies by territory.
    • TNA: events index, per-event detail (example Bound for Glory), and dated preview/news articles. Details provide date, venue, city, local/ET time and sometimes separate doors/countdown/PPV times; previews carry announced bouts/stipulations. AMC/AMC+ rights announcement covers the U.S. weekly show, not every market/event.
    • NJPW: English schedule lists individual dates, cities/venues, and separate door/bell times; some dates link to a “Match Lineup.” Dated official preview example lists bouts. Some English schedule times have no explicit timezone, so preserve them as source-local/zone-unknown until independently resolved.
    • ROH: events index was sparse in the reviewed page; the homepage highlights featured events, while official news provides event/match announcements. Event/detail and newsroom coverage may need to be joined by title/date. Current per-event HonorClub territory availability was not verified.
    • MLW: upcoming-event category and event table expose dates/locations; individual event/ticket notices add venue and bell/door times, while separate news posts announce cards and commonly warn talent/matches may change. The VEEPS partnership announcement and current event notices describe particular worldwide offerings; do not extrapolate those rights to every MLW program.

    Access and evidence limits

    These are verified promotion-owned public pages and field examples, not verified automated-source contracts. I found no public API documentation in the official pages reviewed; API/feed availability remains undocumented, not disproven. I did not establish permission for automated fetching, caching/retention, or redistribution, nor complete market-specific rights for each event. Before production automation, confirm a permitted API/feed or written/terms-supported page-ingestion path, cadence/rate limits, retention/display rights, and territory rights. Do not promise an API or scrape provider pages absent that review.

    Bounded acceptance for AEW, then WWE

    • Ingest official event and dated card assertions independently of XMLTV. Prove discovery with both missing XMLTV and deliberately stale/wrong XMLTV; preserve the official event/card evidence with URL, publication/fetch time and field-level provenance instead of dropping it or trusting bad guide data.
    • Keep event identity separate from airing/offering identity. Store match participants and exact announced stipulation as versioned assertions linked to the event; show “announced” only from an explicit promotion-owned assertion. Exclude rumors from confirmed cards.
    • Preserve date/time source text. Keep doors, venue/bell time and TV/stream/PPV start as distinct clocks; record explicit timezone when provided and leave it unknown rather than infer one. Do not treat venue start as broadcast start.
    • Recheck dated official updates through the event window; append card replacements, status changes, postponements and cancellations with provenance. A vanished page is not proof of cancellation; require an explicit official notice/status.
    • Do not infer a channel, stream, region, subscription entitlement, or playability from event/card pages. Link an event to an offering only when an authorized source confirms the exact provider/service, market and window; otherwise surface the event with availability unresolved and no fabricated Watch target.
    • Canonicalize/match the event once across promotion, guide and mirror observations; retain distinct verified offerings as alternatives. Do not silently double-count the same event because multiple feeds/pages carry it.

    The AEW slice is complete only when these missing/bad-guide, revision, timezone, cancellation, dedupe and unresolved-availability cases are covered. WWE follows; TNA, NJPW, ROH and MLW remain required promotion coverage.

  8. Jumpstile commented on Oct 3, 2026

    @Jumpstile
    OwnerAuthor

    Wrestling event/card v2 proposal — decision packet

    Draft proposal artifacts are in PR #204. The design remains synthetic-only and unapproved: v2 links card revisions to one unchanged v1 event assertion in the existing canonical event truth layer. V1 schema/converter stay closed; no production adapter, consumer, or UI behavior is authorized.

    Independent technical recommendations (advisory, not architecture approval)

    The current decision owner is the Product Owner / Engineering Manager; existing #191 owner dispositions identify @Jumpstile in that role. This issue currently has no assignee, and the repository record names no independent architecture reviewer. Assign that reviewer through the architecture-proposal workflow. These recommendations do not constitute architecture acceptance or product-owner approval.

    Source gate and next slice

    The #29 source-access gate remains open. #191 research establishes candidate public first-party evidence pages and an MCP access avenue, not automated-use, caching, retention, display, or territory rights. No live source was queried. After formal architecture acceptance and source rights, the next bounded slice is one permitted AEW schedule-window adapter under #29: ordinary v1 event assertion plus one linked v2 card observation per revision, report-only Guide Intelligence, offline synthetic tests. No #30 consumer or #191 UI work in that slice.

    Related to #29, #30, and #36. No closure or approval is claimed.

  9. Jumpstile commented on Oct 3, 2026

    @Jumpstile
    OwnerAuthor

    Source-evidence cross-reference for the #191 wrestling discovery gate: the parent-provided 2026-10-03 community-source research, exact source/document URLs, sample findings, licensing limits, and bounded adapter direction are recorded once in #29 comment 5964911791. For #191, Wikipedia text is only separately attributed community evidence, not promotion-approved truth or complete card coverage; it does not close first-party source rights, event coverage, or card-v2 approval. No live source was queried in this owner session.

  10. Jumpstile commented on Oct 3, 2026

    @Jumpstile
    OwnerAuthor

    Owner direction — personal, continuously researched guide (2026-10-03)

    This records the owner's explicit Oct 3 direction, including “Treat every individual sports, ppv, etc channel as a little research project” and confirmation of the interpretation below. This is required product scope, not merely an ecosystem suggestion.

    Required behavior

    • Each person runs ChannelForge for their own personal guide. Do not assume a centrally hosted commercial republication service.
    • Periodically research/enrich programming across all supported categories, including NFL, NBA, MLB, NHL, other sports, wrestling/PPV/PLE, TV, movies and news. The list is illustrative, not a restriction to these leagues/categories.
    • Use XMLTV programmes and the actual lineup to identify relevant entities, channels, time windows and sources. Also discover missing schedules/events and reconcile stale or wrong guides: XMLTV is an input, not unquestionable truth.
    • Maintain a research/evidence profile for each individual sports, PPV, event and other relevant channel: stable identity and aliases, known carriage/market, programming patterns, authoritative/supplemental sources, upcoming events, guide gaps, observation/freshness times, confidence and unresolved mappings.
    • Revisit profiles as programming changes. Apply content-sensitive cadence, caching/conditional requests, incremental change detection, rate limits and last-known-good behavior. Research must not block interactive browsing.
    • Reuse shared source/event evidence across channels rather than repeat identical research. Keep per-channel identity, territory and airing/offering context.
    • Correlate the same programme despite different names/abbreviations, using fuzzy matching plus corroborating event/episode/participant/time evidence. Learn contextual aliases from confirmed matches. Preserve false-merge safeguards and reversible history.
    • Official sources are preferred for confirmed announcements when access/use is permitted; community sources remain explicitly attributed supplemental evidence. Personal use does not waive source-specific access restrictions.

    Acceptance examples

    1. A PPV channel has no useful XMLTV: the system discovers a supported event, dates/card evidence and uncertainty without inventing carriage or a Watch target.
    2. Two channels use different names for the same supported event: compatible evidence links the event and shares enrichment while retaining separate offerings.
    3. Similar names for different episodes, games, live/replay airings or territories do not silently collapse.
    4. A source outage or ambiguous change preserves last-known-good evidence and surfaces material uncertainty; missing text alone is not cancellation.
    5. Fixed input/evidence/knowledge snapshots produce deterministic decisions; confirmed corrections are reused only within proven scope.

    The owner approved the offline, text-only Wikipedia supplemental development direction with revision/time/attribution/uncertainty, and explicitly instructed implementation of cross-channel fuzzy matching/learned aliases. This is not approval to merge PR #204, activate live production ingestion, purchase sources, create credentials or bypass source restrictions.

    Routing and sequencing

    Reuse existing modules/contracts and prior implementations before adding code: #29 source/event evidence; #30 programme/event matching; #31 learned aliases/history; #35 refresh/LKG operations; #36 storage; #190 delivery plan. Keep PR #204's architecture review boundary explicit. Do not create a monolithic scraper or one agent per channel. Phase implementation against the existing deadlines without treating remaining category coverage as optional.

    Existing #191 labels (type:feature, priority:high, component:ux) remain appropriate; no arbitrary milestone was assigned to this cross-cutting requirement. Component slices retain their established milestones and dependencies.

  11. 6 remaining items

  12. Jumpstile commented on Nov 13, 2026

    @Jumpstile
    OwnerAuthor

    Logo plan for this issue is now merged to main in three PRs:

    CI on #219 passed in full (quality-gates, secret-scan, CodeQL, Gitleaks, Semgrep, SBOM, Windows security tests). Before merge, the lane checks of the PR 3 and fingerprint files ran 0 failures (7 test files for PR 3, 20 test files for the fingerprint work).

    Not verified: a full Build-Lineup run end to end with the new setting, and the whole local suite outside CI. Leaving this issue open until a real lineup build shows the logos.
    🤖 Generated with Claude Code

    https://claude.ai/code/session_01NzHAckAcs78Lu3rv8Gqgrf

  13. Jumpstile commented on Nov 17, 2026

    @Jumpstile
    OwnerAuthor

    Progress on Phase 1 (guide truth), 2026-11-17:

    No release or tag.

  14. Jumpstile commented on Nov 17, 2026

    @Jumpstile
    OwnerAuthor

    Real-lineup scorecard result after the two fixes (read-only run on copies of the real 404-channel lineup; 36,068 programmes from one guide; no URLs or provider details copied). Overall: Review.

    • Accuracy: NotMeasured (one guide, nothing to compare; the old false Pass is gone).
    • Completeness: Review (62 of 404 playlist channels have no programmes; matches an independent count).
    • Freshness: Pass (guide runs to Nov 27).
    • Deduplication: NotMeasured (no programme carries a canonical event id).
    • Classification: Review (high 91, ambiguous 0, unknown 142 of 404; unknown is mostly provider groups with no category rule: PPV, Music, adult).
    • Availability: NotMeasured (readiness cannot run on real data yet).

    Gap found: no existing code turns a real playlist plus guide into the evidence records that the guide readiness check needs (the evidence builder is only called by tests), so Availability stays blind and Accuracy cannot see contradictions. Next slice: a small step that builds that evidence from existing data, then re-run. Not verified: whether the 142 unknown and 62 no-programme channels are expected.

  15. Jumpstile commented on Nov 17, 2026

    @Jumpstile
    OwnerAuthor

    Evidence builder merged (PR 274): guide readiness now runs on a real playlist plus one guide. Real 404-channel lineup copy: 332 channels safe candidates, 72 unresolved (62 have no programmes, 10 have programmes but a tvg-id the existing builder rejects). Availability goes from NotMeasured to Review. Accuracy stays NotMeasured with one guide. Next: give those 10 a stable hashed reference in the mapper (no change to the frozen builder), then re-run.

  16. Jumpstile commented on Nov 17, 2026

    @Jumpstile
    OwnerAuthor

    Merged: ID fix for the evidence step (PR 275). On the real 404-channel lineup copy, the 10 channels whose tvg-id has spaces or symbols but that do have programmes are no longer counted as missing (read-only run; raw ids not written anywhere). Still unresolved are the channels that truly have no programmes (62 in the earlier run). Availability stays Review for that reason. Next: developer docs for the scorecard and the evidence step. Open for #191: a look at the 62 channels with no guide data, and the 142 unknown categories (mostly PPV, music and adult groups).

  17. Jumpstile commented on Nov 17, 2026

    @Jumpstile
    OwnerAuthor

    Docs merged: developer guide section for the scorecard and the evidence step.

    Read-only look at why 62 of 404 playlist channels have no guide data (copies of the real lineup; no URLs or provider details copied):

    • All 62 are declared in the guide's channel list but have no programme rows. None are undeclared, and there are no near-miss id matches (case, spaces, suffix), so this is not an id problem.
    • By kind: adult group 30, music or radio streams 28, pay-per-view placeholders that are empty between events 3, one regular sports channel 1. Why the adult and music groups have no programmes is inference (no guide source expected), not verified.
    • 58 of the 62 are also among the 142 channels with no category.
    • The 142 with no category by group: pay-per-view 69, music 31, adult 30, local network stations 12. The existing category vocabulary has no key for pay-per-view, music or adult, so adding rules there means new vocabulary.

    Not verified: whether the 3 pay-per-view gaps and the one sports gap come from the guide source or from ChannelForge.

    Owner decisions pending: (1) count adult and music groups as no-guide-expected so the 62 shrinks to about 4; (2) whether to add Music, Pay-per-view and Adult category keys (new vocabulary).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    component:uxGUI, review workflows, previews, and user intakepriority:highShould be next; significant impacttype:featureA wholly new capability

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions