Web app (web/)
A React 19 and TypeScript single-page app built with Vite and served by nginx on port 5173. It talks only to the Taxonomy API (VITE_API_URL, default http://localhost:8080); GRE services are reached through the /gre/... gateway, never directly.
Pages
| Route | Menu label | Visible to | Backed by |
|---|---|---|---|
/ | Dashboard | all | /dashboard/overview |
/games, /games/:id | Games Catalogue | all (edit: editor) | /games, /games/{id}, /games/{id}/characteristics, documents, AI extract |
/taxonomy | Taxonomy Model | all (edit: editor) | /taxonomy/* |
/similarity | Similar Games | all | /games/{id}/similar, /compare, /similarity/map, /similarity/heatmap |
/engine | Similarity Engine | editor | /similarity/settings, /similarity/settings/preview |
/simulation | GRE Simulation | all (actions: editor) | /gre/*, /hubs/pipeline |
/tagging | AI Tagging | editor | /tagging/*, /review/tags/*, /conflicts |
/settings | Preferences | all | local preferences |
/settings/capture, /capture | Page capture | editor | /games/{id}/documents (a bookmarklet sends the page HTML from the browser when a site blocks automated downloads) |
/settings/providers | AI providers | admin | /ai/providers |
/settings/kafka | Kafka ingestion | admin | /gre/ingest/sources |
/settings/ai-tagging-agent | AI Tagging Agent | editor | /ai/prompts/tagging |
/users | Users | admin | /users |
/connectivity | Connectivity | all | How to connect AI clients (Claude, Claude Code, Cursor) to the MCP server |
Simulation page
web/src/pages/Simulation.tsx is a workspace built around a clickable pipeline diagram (web/src/sim/PipelineFlow.tsx). Each box is a step; selecting it opens that step's panel below the diagram (?step= in the URL). The boxes are colour-coded and their connectors animate with live traffic from the SignalR snapshots.
| Step box | Panel components | What it does |
|---|---|---|
generate, store, profiles | DataTab | Generate a simulated world (batch run), see what the activity service holds, build affinities |
stream, kafka | LiveTab | Start a stream to a chosen ingestion; throughput, lag and Kafka state |
profile | LiveTab | Live Redis profiles of streaming customers |
recommender | LiveWatch, MapTab (live), DataTab (crowd), TuneTab | Watch one customer's list change live, the co-play network, crowd builds, and weight tuning with saved vs draft lists |
similarity | DataTab (similarity), MapTab | Catalogue build and the tag-similarity network |
output | CustomersTab, WhatIfTab, EvaluateTab | Three views: one customer's story (activity, affinities, recommendations; stars mark ground-truth favourites), a hand-built what-if customer, and persona fit plus the holdout accuracy test (AccuracyCard) |
how | HowTab | Plain-language explanation (also the How it works button) |
Reset data (ResetData) calls DELETE /data on the services to start a simulation clean.
The colour of each recommendation bar segment encodes the clue: green for tagged alike, blue for played together, orange for popular, purple for new.