Martha Stewart TV Archive

An unofficial, fan-made archive of Martha Stewart's television work — 2,842 episodes across 12 shows (1986–2024), rendered in a 1990s Martha Stewart Living visual identity, with cross-referenced hubs for the people who appeared on the shows and the places Martha visited.

Live: https://martha.fly.dev

Not affiliated. This project has no connection to Martha Stewart, Martha Stewart Living Omnimedia, Marquee Brands, or any broadcaster. It is a non-commercial fan archive of broadcast facts — episode metadata, segment titles, guest and location names. It hosts no video, no transcripts, and no magazine content. If you are a rights-holder with a concern, open an issue and we will respond promptly.

What's inside

• Episode archive — faceted browse/search over 2,842 episodes (/episodes), per-show pages, a broadcast calendar, and /gaps (an honest accounting of what's still missing).
• People (/people) — 110+ named chefs, contributors, family members, and guests from Martha Stewart Living TV plus guests from the later shows, each with researched bios and episode appearances.
• Places (/places, /places/map) — 340+ farms, bakeries, museums, gardens, restaurants, and field-trip destinations, browsable and plotted on a vintage-atlas map.
• Entity pipeline — scripts/mst-extract-entities.mjs builds the people/places dataset from curated allowlists, researched manual credits, and an LLM-verified extraction over episode segment rundowns (data/marthastewart-tv/segments-llm.json, with verbatim evidence lines and an audit trail in data/marthastewart-tv/llm-review/).

Stack

• Hono + TypeScript JSX, fully server-rendered
• Drizzle ORM + Postgres (Neon in production)
• fly.io (fly.toml, app martha)
• Self-hosted Cormorant Garamond / Libre Caslon Text / EB Garamond

Run locally

pnpm install

# any Postgres works; for a throwaway one:
docker run -d --name martha-pg -e POSTGRES_PASSWORD=martha \
  -e POSTGRES_DB=neondb -p 5433:5432 postgres:16
# .env: DATABASE_URL / DATABASE_URL_UNPOOLED -> your Postgres

pnpm db:migrate
pnpm data:import       # seeds episodes from data/ JSON
pnpm dev               # http://localhost:8080

.env keys: DATABASE_URL, DATABASE_URL_UNPOOLED, SESSION_SECRET, ADMIN_TOKEN, RUN_MIGRATIONS=1, DATA_DIR=./data (see .env.example).

A note on data completeness

This public repo ships the code, the CC-BY episode dataset, and the derived entity dataset. A fresh clone is close to — but not identical to — the live site. Concretely, after pnpm data:import you get:

• 2,497 episodes (the live site shows 2,842 — the extra ~345 are Hallmark-era rows reconstructed from the marthastewart.tv catalog crawl, which is not redistributed; scripts/mst-augment.mjs re-adds them if you have your own crawl).
• /people and /places populate once you additionally run node scripts/mst-persist-entities.mjs (reads the committed data/marthastewart-tv/entities.json). Entity → episode appearance links need the crawl-derived episodes.mst_vhx_id mapping, so on a fresh clone the hubs show entities and bios but few episode cross-links.
• No episode thumbnails (public/episode-images/) — stills are fetched at deploy time from the streaming service's public CDN by scripts/mst-download-thumbs.mjs; we don't republish them in git.

The raw marthastewart.tv catalog crawl (data/marthastewart-tv/raw/, videos.json, items.json, …) requires your own marthastewart.tv subscription — see scripts/mst-crawl.mjs and docs/data-sources.md.

Extending the people/places dataset

• New recurring contributor → add to PEOPLE in scripts/mst-extract-entities.mjs (aliases + researched role).
• New researched business → add to CURATED_PLACES with a role.
• Chef known only by signature dish → add a MANUAL_CREDITS entry.
• False positives → NAME_STOP / PLACE_STOP sets.

Note: re-running the extractor requires the non-redistributed crawl (data/marthastewart-tv/items.json + raw/), so outside contributors should submit allowlist edits by PR without re-running — a maintainer regenerates entities.json and re-persists. node scripts/mst-persist-entities.mjs (which loads the committed entities.json into Postgres) works on any clone.

Licensing

• Code: MIT
• Data (data/*.json, descriptive prose): CC BY 4.0. Bare facts (titles, dates, names) are not copyrightable and are free to reuse regardless.

Deploy

fly deploy --remote-only --app martha

CI (.github/workflows/ci.yml): typecheck + unit tests + build on push/PR.