A graph of TF2 competitive players, built from public match logs. Each node is a player; each edge means they appeared in the same log. Edge weight = number of shared logs; same-team count distinguishes teammates from opponents.

Rewrite of an older project (kyleflett/logMap, ~2021) that used Python + Gephi to render a static 14000×14000 PNG. This version is a trends.tf-fed pipeline into SQLite plus an interactive sigma.js frontend.

pipeline/    # Python, uv-managed. Discovers + fetches logs, exports graph JSON.
frontend/    # Vite + TS + sigma.js static site that renders the graph.

The seam between them is a graph JSON file. Schema: frontend/src/graph/schema.ts.

# 1. Pull logs into SQLite (one-time, ~5 min for a season-sized slice)
cd pipeline
uv sync
uv run logmap discover --format sixes --league rgl --from 2021-05-17 --to 2021-08-15
uv run logmap fetch
uv run logmap resolve-names
uv run logmap status

# 2. Export to JSON the frontend can render
uv run logmap export \
  --format sixes --league rgl \
  --from 2021-05-17 --to 2021-08-15 \
  --min-logs 5 \
  --out ../frontend/public/sample-graph.json

# 3. Run the frontend
cd ../frontend
npm install
npm run dev    # http://localhost:5173/

Pass ?graph=<url> to load a different graph JSON without rebuilding.

• trends.tf is rate-limited to 10 req/min site-wide. The client enforces this globally; do not bypass.
• logs.tf has no documented limit; the fetcher runs 5 concurrent workers conservatively.
• Idempotent + resumable. Re-running discover or fetch skips work already done.
• Raw log JSON is stored zstd-compressed in log_raw so re-shredding is offline.

• trends.tf — log discovery (search by format, league, date range).
• logs.tf — per-log JSON (players, classes, kills/deaths/damage, names).
• RGL via [unofficial RGL API] — team affiliations (planned, not yet wired in).

Phase 1 (data pipeline) and Phase 2 (aggregator) work end-to-end against real data. Phase 3 (frontend) renders. Phase 4 (name resolution from raw logs) is in.

Not yet:

• Wider discovery (the original goal mentioned decades of logs going back to ~2012 — current slice is one RGL season).
• RGL team-membership enrichment.
• logs.map / scores backfill in the logs table (data is in log_raw but never re-shredded).
• Real frontend polish: 42k edges in a season is a hairball without edge bundling, smarter alpha tuning, or higher default --min-logs.

The 2021 version lives at https://github.com/kyleflett/logMap. Kept for historical reference. Not maintained.