Company · Data Model

What this model represents

A portfolio company — an operating company a VC firm has invested in (or is referenced in the deck as a track-record / exemplar / competitor). One .md file per company; sibling-colocated brand assets following the trademark__{TrainCase}.{ext} + favicon__{TrainCase}.{ext} naming convention.

The CEO of the company is a Person (see Person-Data-Model.md) — stored either next to the company file (humain pattern: {co-slug}-ceo.md) or referenced via a ceo_slug field (calmstorm pattern).

Where it lives in the filesystem

Naming convention: kebab-case slug; .md extension.

Brand assets and discovery globs

The shell discovers companies via .md filenames; assets are discovered by image-extension glob, independent of any registry. Asset filename convention:

{Company-Name} is Train-Case (Foundation-Health, Unnatural-Products, 9am-Health), NOT kebab. This is a human-readability convention; the shell’s asset glob doesn’t care:

// apps/deck-shell/src/routes/data-assets/companies.astro
const assetGlob = import.meta.glob(
  "/data/**/portfolio/*.{png,jpg,jpeg,webp,avif,svg,ico}",
  { eager: true, query: "?url", import: "default" }
);

The audit route renders whatever’s there — Train-Case, kebab, mismatched, doesn’t matter. Don’t enforce naming in the DB; enforce by convention.

Asset strategy metadata (chroma+calmstorm only)

Chroma and calmstorm record which tier of the fetch cascade succeeded in frontmatter — useful for the future “re-fetch all rasters as SVG” pass:

logo: ./trademark__9am-Health.png
logo_format: png
logo_source: https://...
logo_bg_color_detected: "#000000"
logo_bg_strip_method: imagemagick-floodfill
logo_bg_stripped: true
logo_asset_strategy: og-image-stripped   # ← which tier succeeded
favicon: ./favicon__9am-Health.png
favicon_format: png
favicon_strategy: site-link-icon
og_image_url: https://...
og_image_strategy: og-image

Possible logo_asset_strategy values observed: site-svg-path, site-raster, site-raster-stripped, og-image, og-image-stripped, brandfetch-svg, wikimedia-svg.

Humain omits all of these — the humain ingest didn’t run the full asset cascade (just grabbed the OG image as WebP from Squarespace). That’s a content gap, not a schema gap. The DB column should be nullable.

The canonical schema (current — chroma + humain convention)

---
slug: unnatural-products             # required; kebab-case; filename stem
name: "Unnatural Products, Inc."     # required; display name
short_name: "UNP"                    # optional; abbreviation
website: https://www.unnaturalproducts.com  # required; homepage
homepage: https://www.unnaturalproducts.com # synonym; calmstorm uses `homepage`, chroma uses `website` ALSO

# Categorization
sector: "TechBio · Drug discovery"
sub_sector: "Synthetic macrocyclic peptide therapeutics"
stage: "Series A / partnership-driven"
location: "Santa Cruz, California, USA"
founded: 2018

# Firm linkage
firm_slug:                           # FK to Firm.slug; SET for chroma's backer-portfolio companies; NULL for humain's own portfolio

# People linkage
founders:
  - "Cameron Pye, PhD"
  - "Joshua Schwochert, PhD"
ceo: "Cameron Pye, PhD"
ceo_file: unnatural-products-ceo.md  # humain convention: sibling Person file
ceo_slug: unnatural-products-ceo     # calmstorm convention: FK to Person.slug

# Engagement-specific
humain_relationship: "Active Fund I portfolio (first warehoused investment, $40M post; pro-rata rights)"
deal_validation:
  - partner: "BridgeBio"
    year: 2021
  - partner: "Merck"
    year: 2023
    note: "Humain invested at $40M post around this milestone"
aggregate_deal_value: "$2B+ (per UC Santa Cruz Dean's Council profile, 2025)"
equity_raised: "$38M+"

# Body
description_short: "AI-designed macrocyclic peptide therapeutics — bridging the gap between small molecules and biologics"

# Brand assets
trademark: trademark__Unnatural-Products.webp
trademark_source: "static1.squarespace.com (UNP_Logo_Full_1200-01.png served as WebP)"
favicon: favicon__Unnatural-Products.webp
og_image_url: http://static1.squarespace.com/static/...
asset_strategy: site-raster

# Provenance + lifecycle
confidence: high
ingested_at: 2026-06-06
source_urls:
  - "https://www.unnaturalproducts.com"
  - "https://www.unnaturalproducts.com/about"
review_status: pending               # chroma-only; pending | reviewed | needs-rework
notes: "free-form authoring notes"
---

# Free-form markdown body — thesis, platform, partnerships, investment context.

Older / calmstorm shape — preserved for backward compat

---
slug: 9am-health
name: "9am.health"
homepage: https://www.9am.health/    # NOTE: `homepage` not `website`
firm_slug: calm-storm-ventures

# Asset metadata fully populated
logo: ./trademark__9am-Health.png    # NOTE: `./` prefix; `logo` field name
logo_format: png
logo_source: https://...
logo_bg_color_detected: "#000000"
logo_bg_strip_method: imagemagick-floodfill
logo_bg_stripped: true
logo_asset_strategy: og-image-stripped

description: "..."
sector: "Health Frontend, Health Backend"   # comma-separated string
deck_name: "9am Health"                     # calmstorm-only; verbatim deck label
deck_category: "Chronic Conditions"         # calmstorm-only

profiles:                            # NESTED ARRAY (calmstorm only)
  - type: firm-bio
    url: https://...
  - type: homepage
    url: https://www.9am.health/

ceo_slug: 9am-health-ceo

# Provenance
sources: [...]                       # NOTE: `sources` not `source_urls`
fetched_at: 2026-05-10T18:56:33Z     # ISO datetime
confidence: high
status: complete                     # NOTE: `complete` not `resolved`

# Favicon — same shape
favicon: ./favicon__9am-Health.png
favicon_format: png
favicon_strategy: site-link-icon
favicon_source: https://...
og_image_url: https://...
og_image_strategy: og-image
---

Per-client variations (and why)

Why the variation: same answer as Person — calmstorm was first, chroma rewrote, humain inherited chroma’s shape. The humain_relationship is engagement-specific narrative (one fund’s-eye view); the DB shouldn’t try to schematize it strictly — carry as JSON.

How the shell consumes Company rows

Route: apps/deck-shell/src/routes/data-assets/companies.astro

Discovery glob: import.meta.glob("/data/**/portfolio/*.md", { eager: true, query: "?raw", import: "default" })

Filter applied: rows where frontmatter.role_class is NOT set. (Persons — see Person-Data-Model.md — also live under portfolio/, distinguished by the presence of role_class.)

Fields the route reads:

name · slug · homepage / website · firm_slug
sector · description / description_short
logo / trademark · favicon · og_image_url
asset_strategy fields (logo_asset_strategy, favicon_strategy, …)
confidence · status · ingested_at / fetched_at
ceo_slug (if present)

The audit route also pairs each company row with its colocated brand-asset files via the parallel asset glob.

What’s load-bearing

• slug — primary key; filename stem; URL fragment
• Absence of role_class — discriminator from Person on the same portfolio/ glob
• trademark + favicon filenames + colocated files — without these the audit row shows dashes for assets
• firm_slug — required for chroma’s multi-investor pattern (so the audit page can group by firm); nullable for humain’s flat pattern

Translation to a remote DB

Prisma schema sketch

model Company {
  // Identity
  id          String   @id @default(cuid())
  slug        String                          // kebab-case; per-engagement unique
  engagement  String                          // calmstorm-decks | chroma-decks | humain-vc-decks

  // Core
  name        String
  short_name  String?
  website     String?
  description String?  @db.Text

  // Categorization
  sector      String?
  sub_sector  String?
  stage       String?
  location    String?
  founded     Int?

  // Firm linkage
  firm_slug   String?                         // FK to Firm.slug; null for flat operating-company variant

  // CEO linkage
  ceo_slug    String?                         // FK to Person.slug (where role_class = portco-ceo)

  // Brand assets
  trademark_path        String?               // filename relative to company's .md dir
  trademark_format      String?               // svg | png | webp | …
  trademark_source      String?
  asset_strategy        String?               // see enum below; nullable for skipped clients
  logo_bg_stripped      Boolean?
  logo_bg_color_detected String?
  favicon_path          String?
  favicon_format        String?
  favicon_strategy      String?
  og_image_url          String?
  og_image_strategy     String?

  // Engagement-specific narrative (kept as JSON to avoid per-engagement column proliferation)
  engagement_narrative  Json?                 // e.g. humain_relationship, deal_validation[], aggregate_deal_value

  // Deck-label preservation (calmstorm)
  deck_name     String?
  deck_category String?

  // Calmstorm-shape preserved
  profiles_legacy Json?                       // nested {type, url}[]

  // Provenance + lifecycle
  status        String?                       // resolved | flagged | complete (calmstorm)
  confidence    String?                       // high | medium | low | flagged
  ingested_at   DateTime?                     // unified from `ingested_at` and `fetched_at`
  source_urls   String[]                      // unified
  review_status String?                       // chroma-only; pending | reviewed | needs-rework
  notes         String?  @db.Text

  // Full frontmatter capture
  raw_frontmatter Json?

  @@unique([engagement, slug])
  @@index([engagement, firm_slug])
  @@index([firm_slug])
}

Asset-strategy enum (free-form text, but document the known values)

site-svg-inline     — extracted from <svg> inside the company's homepage HTML
site-svg-path       — found at /logo.svg or similar common path
site-raster         — PNG/JPG from /logo.png or similar
site-raster-stripped — same as above, with background flood-filled to transparency
og-image            — derived from <meta property="og:image"> on homepage
og-image-stripped   — og:image with bg strip
brandfetch-svg      — pulled from Brandfetch API (tier 4 of the cascade)
wikimedia-svg       — pulled from Wikimedia / SVG-repo search (tier 5)
google-svg          — Google Custom Search with fileType=svg (tier 6, last resort)

The full cascade is documented in context-v/agent-skills/crawl-fetch-ingest/SKILL.md § “Trademark / wordmark cascade”.

Open questions for the collaborator

1. Brand-asset binary storage: like Person headshots, do brand assets get an S3/CDN URL only, or also bytes-in-DB? Recommendation: URL only.

2. firm_slug as nullable FK: humain’s data/portfolio/ rows have no firm_slug. Should the DB enforce non-null + auto-populate with the engagement’s own firm? Recommendation: leave nullable; humain’s operating-company variant is a real shape that shouldn’t be forced into firm-anchored.

3. Engagement-specific narrative as JSON: humain has humain_relationship, deal_validation[], aggregate_deal_value, etc. — fields the next client may name differently (e.g., chroma_position, co_invest_partners[]). Recommendation: keep these as a Json column engagement_narrative; columnize only if a query layer needs to filter by them across engagements.

4. De-dup across engagements: if Calmstorm and Humain both invest in Company X, two rows or one? Recommendation: two rows, with a canonical_company_id FK that links if/when curated. Same answer as Person.

5. Asset-strategy enum strictness: new tiers may emerge. Recommendation: free-form text in DB; enum lives in the skill spec.

See also

• Person-Data-Model.md — the sibling shape that also lives in data/**/portfolio/
• Firm-Data-Model.md — the firm container that firm_slug references
• context-v/agent-skills/crawl-fetch-ingest/schema/company.md — the upstream skill’s schema spec
• context-v/agent-skills/crawl-fetch-ingest/SKILL.md § “Trademark / wordmark cascade” — full asset-strategy taxonomy