# Film diary sources

The initial catalogue contains **24 watched** and **120 watchlist** feature films. The site owner explicitly requested arbitrary initial watched selections; no personal viewing dates, ratings or notes have been invented. Those fields start empty and are editable in the authenticated admin panel. The catalogue file is an initial seed; the live site reads Supabase and never uses it to overwrite later edits.

## Metadata and original posters

- Movie identities, titles and premiere years were matched to IMDb public suggestion results, checking exact title/year and unique IMDb IDs.
- Horror / thriller classifications were checked against Stremio Cinemeta metadata and the Wikipedia-derived movie dataset at https://github.com/prust/wikipedia-movie-data . A film must have evidence for at least one of the two genres.
- Actual film posters were obtained from the public Metahub poster CDN linked to those movie IDs. Poster artwork belongs to the respective studios, distributors and artists; no open licence is claimed. No artificial posters or substitute photographs were generated.
- Local previews are proportionally resized WebP images up to 400px wide. Full artwork proportions are preserved. All 144 images were decoded successfully and have distinct SHA-256 hashes.
- `catalogue.json` retains the film information and poster source links. `source-manifest.json` records per-film metadata URLs, original poster URLs, local dimensions and hashes.
- The live page loads posters from this site's own files. A poster uploaded later in the admin panel is stored in the existing private media bucket and its backup; visibility follows the film record.

## Automatic admin imports

- Search: IMDb public title suggestions, `https://v2.sg.media-imdb.com/suggestion/<first-letter>/<query>.json`. Original and Turkish names or IMDb title IDs can be searched; the selected movie ID is stored explicitly.
- Metadata and fallback search: the official [Stremio Cinemeta catalogue](https://github.com/Stremio/stremio-addons/blob/master/docs/tutorial/using-cinemeta.md), `https://v3-cinemeta.strem.io/meta/movie/<imdb-id>.json`. Cinemeta aggregates movie data from multiple providers; this is not a claim that every metadata field comes directly from an official paid IMDb API. Missing scores or durations stay empty. Synopsis language is supplied by the provider.
- Downloaded artwork uses the provider's returned poster URL; for Cinemeta's Metahub small poster, the verified medium variant for that same IMDb ID is used. Automatic imports save an actual WebP copy plus backup under the private film-media path, with the returned source URL and fetch timestamp retained. Artwork ownership remains with its original rights holders.
- An IMDb score is an external metadata snapshot. It is distinct from the couple's personal score and viewing notes. Refreshing metadata is an explicit admin action; existing personal history is preserved.

IMDb/Cinemeta years can reflect a festival premiere rather than a later theatrical release. Source records checked 12 November 2026.

## Live recent-film discovery

On 12 November 2026 the owner asked to clear the initial selection. Its 144 live records were moved to the existing reversible trash; the public archive is empty until the owner adds films. The historical seed/provenance files are not a runtime fallback and are not automatically reapplied.

The daily verification builder discovers titles from Netflix's live Turkish [horror](https://www.netflix.com/tr/browse/genre/8711) and [thriller](https://www.netflix.com/tr/browse/genre/8933) catalogues. It reads each candidate's own Movie JSON-LD and **Seslendirme / Audio** section. Turkish subtitles, original Turkish productions and audio-description-only tracks do not qualify. IMDb suggestions identify the film; Cinemeta verifies its IMDb ID, release year, genre and matching director.

`discovery.json` stores those results with their exact audio source and actual check time. A daily workflow refreshes it; `verified-catalogue.mjs` is an embedded generated copy for cold starts. The public function selects from still-valid verified records immediately and refreshes from the fixed GitHub source hourly in the background. Audio evidence expires after seven days; retaining a result during an outage never advances its proof timestamp. Only released movies in the current and four preceding calendar years qualify (2022–2026 at launch), and the range is rechecked on every request. No personal archive record is created.

The result includes a genuine IMDb link, metadata snapshot, actual poster and the exact Netflix audio-source page/check time. Evidence is provider-specific and does not promise identical audio tracks for every platform or account. Provider failures never substitute an old diary film or an unverified dubbing claim. `recent-hints.mjs` prioritizes previously observed IDs during verification and cannot bypass the live checks. Independent IMDb year checks prevent regional Netflix `dateCreated` dates from making older films appear recent.


## Recommendation poster reliability — 13 November 2026

All nine verified discovery records now include real poster previews. Artwork is downloaded from each record's documented `movie.poster_path`, decoded and proportionally resized to WebP, then embedded in `poster` alongside the metadata. It is the actual film artwork, not a generated substitute. Original provider URLs and audio proof dates remain unchanged. Daily verification preserves valid artwork for the same IMDb ID and only publishes recommendations with an intact preview. Recommendations do not create or change personal viewing records.


## Broad IMDb discovery — 13 November 2026

The current picker uses the official [IMDb non-commercial dataset](https://data.imdb.com/non-commercial-datasets/) at `https://datasets.imdbws.com/title.basics.tsv.gz`. Only the complete set of non-adult `movie`/`tvMovie` IDs is retained in a compact checked binary index (908,623 at first build). No genre, year, rating, language or Netflix requirement is applied. The previous nine audio-verified titles are historical and are not the current recommendation pool.

IMDb public title suggestions resolve a selected identity and poster; optional additional fields are from the official Cinemeta IMDb-ID metadata service. Existing artwork rules and original rights holders remain unchanged. Poster files are transferred with the result after safe source/redirect and image signature validation. No unwarranted dubbing or availability claim accompanies broad results.

Information courtesy of IMDb (https://www.imdb.com). Used with permission.


Optimized copies of the currently public MetaHub posters are listed in [PUBLIC_MEDIA_SOURCES.json](../PUBLIC_MEDIA_SOURCES.json). Poster rights remain with the respective rights holders.
