How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

draft-hub-performance - Skill

Name: draft-hub-performance
Author: RandomStatsGuy

Instruction file imported from RandomStatsGuy/ScoreSense (.cursor/rules/draft-hub-performance.mdc). Copyright stays with the author.

Fantasy performance

Prefer materialized caches over live predict_* on hub routes: artifacts/draft_pool/, artifacts/weekly_predictions/.
Pool vs overlay: build_draft_pool_payload() then build_value_overlay() in value_sheet.py. Do not recompute the pool per request.
Roster/cap reads are DB-only. Sleeper runs on link/sync, or GET /api/hub/roster?live_sleeper=1 for a one-off pull.
GET /api/hub/home and GET /league/{id}/freshness serve the last-known snapshot (stale_as_of) and never parse commissioner workbooks or block on Sleeper. Cap-sheet stale uses file mtimes vs stored file_fingerprint. Workbook parse stays on explicit sheet sync.
Hub GET /media/{id} serves the original unless ?w= is set. Snap to 48 / 96 / 256, cache the WebP variant next to the file, and never resize on a cache hit. Clients request the painted size — do not ship the studio original for a 22px mark.
Scoring history lives in SQLite sleeper_scoring_cache. Refresh on league sync or GET /api/hub/league/{id}/insights?refresh=1.
CPU-heavy work uses app/process_pool.py. Do not start a per-route process pool.
Secrets and paths go through src/config.py, not raw os.environ in routes.
HUB_TIMING=true logs slow hub routes (>500ms) and may set X-Hub-Timing-MS.
After mlready or model changes, refresh the draft-pool fingerprint (preseason_refresh or scripts/ops/fix_artifact_fingerprints.py). Do not leave first-visit Fantasy pages on cold live inference.

Measurement and perceived speed

Perceived speed follows the same architecture: paint roster/cap snapshots first, retain known content during refresh, reserve image/skeleton geometry, and make selection/previews immediate. Trades loads Ideas/Inbox on demand and reuses the existing account/league/team-scoped weekly snapshot. Measure cold/warm navigation and a representative interaction before claiming an improvement; record device/viewport, cache state, and data source.

Future field-measurement targets are LCP ≤2.5s, INP ≤200ms, and CLS ≤0.1 at the 75th percentile, segmented by phone/desktop (Core Web Vitals thresholds). These are targets, not current results; one local lab render cannot establish a field percentile. Do not add a telemetry dependency or more polling just to populate the rollout log.

Install draft-hub-performance and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

{ "ocm": "1", "id": "randomstatsguy-scoresense-draft-hub-performance-rule", "kind": "skill", "name": "draft-hub-performance", "description": "Fantasy / hub performance — caches, no live ML on hot paths.", "publisher": "RandomStatsGuy", "version": "1.0.0", "capabilities": { "domains": [ "coding", "math" ], "tags": [ "skill-md", "github-cursor-rules" ], "languages": [ "en" ] }, "quality_prior": 0.6, "examples": [ "Fantasy / hub performance — caches, no live ML on hot paths." ], "primary": false, "metadata": { "source": { "provider": "github-cursor-rules", "repository": "https://github.com/RandomStatsGuy/ScoreSense", "path": ".cursor/rules/draft-hub-performance.mdc", "ref": "6e00c2ca4c9f7f55df5a76978db1e5c88c878a64", "url": "https://github.com/RandomStatsGuy/ScoreSense/blob/6e00c2ca4c9f7f55df5a76978db1e5c88c878a64/.cursor/rules/draft-hub-performance.mdc", "key": "RandomStatsGuy/ScoreSense/.cursor/rules/draft-hub-performance.mdc" }, "applies_to": "src/draft_hub/**/*.py, app/hub_routes.py, frontend/src/DraftHub/hubDataCache.js" }, "instructions": "# Fantasy performance\n\n- Prefer materialized caches over live `predict_*` on hub routes: `artifacts/draft_pool/`, `artifacts/weekly_predictions/`.\n- Pool vs overlay: `build_draft_pool_payload()` then `build_value_overlay()` in `value_sheet.py`. Do not recompute the pool per request.\n- Roster/cap reads are DB-only. Sleeper runs on link/sync, or `GET /api/hub/roster?live_sleeper=1` for a one-off pull.\n- `GET /api/hub/home` and `GET /league/{id}/freshness` serve the last-known snapshot (`stale_as_of`) and never parse commissioner workbooks or block on Sleeper. Cap-sheet stale uses file mtimes vs ", "cost": { "context_tokens": 593 } }

draft-hub-performance

About

Fantasy performance

Measurement and perceived speed

Install draft-hub-performance and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews