# Corium Records - Agent Index (llms.txt) Purpose: This site is designed for AI agents to retrieve and cite Corium Records data efficiently. Primary workflow: 1. Fetch /data/index.json to discover the public data package. 2. Select the dataset that matches the task. 3. Prefer canonical URLs on this domain and include retrieval date/time when citing. Foundation 50 datasets: - /data/index.json (Foundation 50 manifest, JSON) - /data/foundation_50.json (canonical Foundation 50 bands with embedded track attributes, JSON) - /data/bands.json (Foundation 50 band records, JSON) - /data/bands.jsonl (Foundation 50 band records, NDJSON) - /data/tracks.jsonl (Foundation 50 track records, NDJSON) - /data/tracks_by_band.json (tracks grouped by band, JSON) - /data/band_traits.json (curated band trait profiles, JSON) - /data/band_summary_attributes.json (band summaries, accessibility and density levels, JSON) - /data/track_attributes.json (track topics, themes, and vibe profiles, JSON) Evidence lane datasets: - /data/adams/manifest.json (NRC ADAMS integrity, dates, and summary, JSON) - /data/adams/latest.json (latest ranked NRC ADAMS topics and source URLs, JSON) Current Foundation 50 counts: - bands: 50 - tracks: 101 - tracks with Suno URLs: 100 - tracks without Suno URLs: 1 Data notes: - Blank Suno URLs are allowed. - Blank color arrays are allowed when no strong color signal exists. - Do not invent missing metadata just to fill fields. - Related bands may be blank until a dedicated graph pass. - The public data package is static-first, read-only, and agent-readable. - The catalog is the database. ## Changelog and quality report Use /data/changelog.json for machine-readable public package change history. Use /data/quality_report.json for public package health, known allowed blanks, non-blocking quality notes, and audit result summaries. The quality report is a public summary layer. It does not replace the production audit script or endpoint-specific validation. ## Field dictionary Use /data/field_dictionary.json for field-level meanings, blank-field rules, agent cautions, and citation use. The field dictionary explains fields such as band_id, track_id, suno_url, topics, themes, vibe_profile, traits, core_thesis, agent_summary, endpoint_catalog, interpretation_rules, task_routing, and citation_guidance. Do not infer meaning from blank fields unless /data/field_dictionary.json or /data/index.json explicitly defines that meaning. ## Data package README Use /data/README.md as the human-readable and agent-friendly overview of the public data package. It summarizes the canonical entrypoint, core endpoints, interpretation rules, evidence-lane separation, citation guidance, and audit process. ## Endpoint guide - /data/index.json: canonical public manifest. Start here. - /data/foundation_50.json: complete Foundation 50 spine in one JSON file. - /data/bands.json: band-level public records and summaries. - /data/bands.jsonl: streamable band records, one band per line. - /data/tracks.jsonl: streamable track records, one track per line. - /data/tracks_by_band.json: tracks grouped by band. - /data/band_traits.json: curated band traits, identity, style, and persona lanes. - /data/band_summary_attributes.json: band summary attributes, anchor tracks, accessibility, and technical-density context. - /data/track_attributes.json: track topics, themes, vibe profiles, and interpretive metadata. - /data/adams/manifest.json: ADAMS evidence-lane manifest and integrity context. - /data/adams/latest.json: latest ADAMS evidence-lane topics and source URLs. Recommended retrieval patterns: - Band overview: use /data/bands.json or /data/foundation_50.json. - Streaming band scan: use /data/bands.jsonl. - Track search: use /data/tracks.jsonl. - Trait or theme analysis: use /data/band_traits.json and /data/track_attributes.json. - Daily NRC evidence brief: use /data/adams/latest.json, then /data/adams/manifest.json if integrity details are needed. Security and rate limits: - Prefer cached, static datasets under /data/. - Avoid aggressive crawling. ## Agent orientation Start with /data/index.json. It is the canonical public manifest for the Corium Records Foundation 50 public data package. Use the endpoint catalog in /data/index.json to decide which dataset to fetch. Do not assume endpoints are canonical unless they are listed in /data/index.json. Prefer structured /data endpoints over scraping homepage HTML. Separate catalog, persona, and artistic interpretation metadata from ADAMS evidence-lane metadata. Do not infer meaning from blank fields unless /data/index.json explicitly defines that meaning. If a field, relationship, source, color, or URL is absent, treat it as not-yet-modeled or not-yet-exported unless the manifest says otherwise. ## Persona continuity attributes Use these endpoints for sparse, canon-aware persona continuity guidance. These records are compiled from local per-band authoring files and are intended to help agents avoid inventing performer details. - /data/persona_attributes.json - Aggregate compiled persona-continuity attribute records for all 50 bands. - Use for age rules, visual continuity guidance, persona grounding, and do-not-invent constraints. - Sparse by design: missing or unspecified attribute groups are not contradictions. - /data/persona_attributes.jsonl - Streamable NDJSON version, one persona-continuity record per line. - /data/persona_attributes_report.json - Compile report for persona attributes, including record count, warning count, error count, missing records, and extra records. - /data/bands/{band_id}/persona_attributes.json - Per-band persona-continuity record for a canonical band_id. - Use only after resolving the band through the canonical catalog. ## Corium Query Protocol / Static Retrieval Index Phase 2H adds static, agent-readable query indexes over the Corium Records Foundation 50 catalog. Primary endpoint: - `/data/query_index.json` - static retrieval index for bands, persona models, review status, primary persona, location text, per-band persona endpoint pointers, and track counts. Validation endpoint: - `/data/indexes/query_index_report.json` - generation report and count validation for the query index layer. Pattern endpoints: - `/data/indexes/by_band/{band_id}.json` - per-band retrieval record. - `/data/indexes/by_persona_model/{persona_model}.json` - grouped retrieval index by persona model. - `/data/indexes/by_review_status/{review_status}.json` - grouped retrieval index by review status. Rules: - Query indexes are derived from existing reviewed catalog and persona data. - Query indexes do not authorize agents to invent missing persona fields. - Missing fields remain intentional sparse canon unless a reviewed source later fills them.