Enrichment script: fill missing quote fields using TMDB, Open Library, and Wikidata

[DutchJaFO]

Purpose

Define the shared enrichment infrastructure that all provider implementations (#35, #36, #37, #46, and any future providers) build on. Each provider is independently shippable once this scaffold is in place — they do not depend on each other.

What this issue delivers

IEnrichmentProvider interface

public interface IEnrichmentProvider
{
    /// <summary>The source types this provider handles (e.g. movie, tv, book, person).</summary>
    IReadOnlySet<string> SupportedTypes { get; }

    /// <summary>Attempt to fill missing fields on the quote. Only populates null/empty fields — never overwrites curated data.</summary>
    Task<EnrichmentResult> EnrichAsync(Quote quote, CancellationToken ct = default);
}

EnrichmentResult carries: fields changed, confidence level, and any flags for manual review (fuzzy match, ambiguous result, etc.).

Runner skip rules

The enrichment runner must check two conditions before passing a record to any provider:

1. isComplete: true — skip the entire record; the user has marked it done (Schema: record completeness flag and per-field verified-absent markers #55)
2. noValueKnown — skip individual fields listed in this array; the user has confirmed no value exists for them (Schema: record completeness flag and per-field verified-absent markers #55)

Providers must not attempt to fill a field present in noValueKnown, even if their lookup returns a confident match. These are deliberate human decisions that enrichment must not override.

Cascading / universal scraper pattern

Inspired by Kodi's Universal Movie Scraper: the enrichment runner tries each registered provider in priority order and merges results. A provider that returns a confident match for a field wins that field; a provider that returns no match leaves the field for the next provider to try. No single source is authoritative.

Script scaffold (scripts/enrich.csx)

• --dry-run flag: print proposed changes without writing
• --no-fetch flag: use scripts/cache/ instead of live API calls
• --provider flag: run only one provider (e.g. --provider tmdb) for targeted re-enrichment
• --include-complete flag: override the isComplete skip rule and process all records (useful for re-enrichment after adding a new provider)
• Per-provider cache directories under scripts/cache/<provider>/
• Structured log output: one line per quote, listing fields changed, fields skipped due to noValueKnown, and any manual-review flags
• Provider registration: each provider is added as a step in the runner; adding a new provider requires no changes to existing providers

Where the shared code lives

Quotinator.Enrichment class library, referenced by both scripts/enrich.csx and Quotinator.Api (for the import endpoint's optional enrich pass — see #45). The script is a thin runner over this library.

Notes

• Providers: TMDB (Enrichment script: TMDB (movies and TV) #35), Open Library (Enrichment script: Open Library (books) #36), Wikidata (Enrichment script: Wikidata (people) #37), TVDB (Enrichment: TVDB provider for TV quotes #46), AniList (Enrichment: AniList provider for anime quotes #49)
• Each provider issue depends only on this scaffold — not on each other
• The --provider flag makes it safe to run one enricher before others are implemented
• Depends on: completeness and verified-absent fields (Schema: record completeness flag and per-field verified-absent markers #55) being in the schema before the skip rules can be implemented