Clusterflick is a free, open-source site listing every film showing in London, from 450+ venues in one search: chains, independents, film clubs, festivals and pop-ups. Each film shows its screenings across every venue, with ratings from IMDb, Letterboxd, Rotten Tomatoes and Metacritic, and filters for subtitled, audio-described, relaxed and baby-friendly showings.

🎬 Looking for something to watch? Head to https://clusterflick.com 🍿

This organisation holds the code behind it. Listings are pulled from each venue's own website several times a day, put into one format, matched to The Movie Database, and built into a static site. The pipeline runs in the open as a chain of GitHub Actions, one repository per stage (see Pipeline).

📄 For data you can use, see clusterflick.com/data-licence.

Most of the repositories below are internal build artifacts, published because the pipeline runs in the open rather than offered for reuse. Each one carries a LICENSE-DATA file saying which it is.

The code in every repository is separately available under the MIT licence.

The site

• clusterflick.com — the website
• api-tmdb-search — the Cloudflare Worker behind the site's film search
• branding — brand guide, logos and icons

The data pipeline

• scripts — the code behind every stage, with a module for each venue and ticketing source
• data-retrieved — pulls raw listings from venue websites and ticketing platforms
• data-transformed — puts each venue's listings into one format and matches films to TMDB
• data-diffed — records what changed between runs
• data-backedup — keeps a copy of every transformed release off GitHub, one archive per week
• data-cached — caches TMDB metadata for every film
• data-combined — merges every venue into the single dataset the site is built from
• data-matched — adds ratings from IMDb, Letterboxd, Rotten Tomatoes and Metacritic
• data-calendar — calendar feeds, one per venue
• host-non-web-sources — hand-transcribed listings for venues with nothing on the web to read

Tooling and operations

• data-analysed — finds new venues, checks listings against other sources, tracks pipeline reliability
• manage.clusterflick.com — internal review site for the pipeline
• analysis.clusterflick.com — the original site, kept for reviewing the combined data
• generate-spotlights — social media posts
• self-hosted-workflows — workflows for the Raspberry Pi runners
• themoviedb-data — TMDB catalogue download for experimentation
• release-downloader — fork of a GitHub Action for downloading release assets
• .github — this profile and shared reusable workflows

1. 
2. 
   • (only releases when something changed)
     • (runs whether or not the diff released)
   • (archives each finished week to Cloudflare R2; also daily)
   • also triggers scripts PR and data-analysed analysis scripts
3. 
   • 
     • (only releases once per day)
   • 
     • clusterflick.com
       • manage.clusterflick.com (also after each venue health cycle, and nightly)
   • analysis.clusterflick.com

flowchart LR
    retrieved["data-retrieved"]
    transformed["data-transformed"]
    calendar["data-calendar"]
    cached["data-cached"]
    diffed["data-diffed"]
    backedup["data-backedup"]
    nonweb["host-non-web-sources"]
    combined["data-combined"]
    matched["data-matched"]
    scripts["scripts"]
    analysed["data-analysed<br>(comparisons)"]
    runstats["data-analysed<br>(run stats)"]
    llmusage["data-analysed<br>(LLM usage log)"]
    releaselog["data-analysed<br>(release log)"]
    venuehealth["data-analysed<br>(venue health, hourly)"]

    subgraph retrival[" "]
        direction TB
        retrieved --> retrival-release>Release]
    end

    subgraph transformation[" "]
        direction TB
        transformed --> transformation-release>Release]
    end

    subgraph calendars[" "]
        direction TB
        calendar --> calendar-release>Release]
    end

    subgraph caching[" "]
        direction TB
        cached --> cached-release>Release]
    end

    subgraph diffing[" "]
        direction TB
        diffed --"if changed"--> diffed-release>Release]
    end

    subgraph backup[" "]
        direction TB
        backedup --"each finished week"--> backup-archive[("R2 archive")]
    end

    subgraph combination[" "]
        direction TB
        combined --> combined-release>Release]
    end

    subgraph matching[" "]
        direction TB
        matched --> matched-release>Release]
    end

    subgraph website[" "]
        direction TB
        clusterflick.com --> id1[[Generate website]]
        analysis["analysis.clusterflick.com"] --> id2[[Generate analysis site]]
        manage["manage.clusterflick.com"] --> id3[[Generate management site]]
    end

    nonweb -.->|read by| retrival
    retrival --triggers--> transformation
    transformation --triggers--> backup
    transformation --triggers--> diffing
    transformation --triggers--> scripts
    transformation --triggers--> analysed
    transformation --triggers--> runstats
    transformation --triggers--> llmusage
    transformation --triggers--> releaselog
    diffing --triggers--> caching
    caching --triggers--> combination
    combination --triggers--> calendars
    combination --triggers--> matching
    combination --triggers--> analysis
    calendars --triggers--> clusterflick.com
    calendars --triggers--> releaselog
    matching --"triggers (if new data)"--> clusterflick.com
    clusterflick.com --triggers--> manage
    venuehealth --triggers--> manage

• scripts
• clusterflick.com
• api-tmdb-search
• data-retrieved (manual — one venue on a chosen runner)

• (daily at 11am UTC)
• (manual)
• (after each transform)
• (after each transform)
• (after each transform)
• (after each transform)
• (after each transform — feeds the stability badges above)

Failed runs of these stages are re-run automatically, up to 3 attempts.

• (after each transform)
• (every 3 hours — catches runs the log missed)
• (after each transform and calendar release)
• (hourly)
• (hourly)
• (weekly)

• 
• 
• 
• (manual)
• (manual — new public IP for every self-hosted runner)

Each repo audits its GitHub Actions workflows with zizmor via the shared .github reusable workflow.

• data-retrieved
• data-transformed
• data-backedup
• data-calendar
• data-cached
• data-diffed
• data-combined
• data-matched
• data-analysed

• clusterflick.com
• analysis.clusterflick.com
• api-tmdb-search

• github
• scripts
• generate-spotlights
• self-hosted-workflows
• release-downloader