Yield/Fidelity: checks whether active Beefy vaults on Base earn their reported APY

Live: https://yield.rahilbhavan.com

Yield/Fidelity is a Next.js dashboard for active Beefy vaults on the Base network. Once a day, it reads each vault's price per share (PPS) on-chain and stores the reading with its block provenance in Supabase. It then compares realized PPS growth with the APY that Beefy reported over the same intervals. It also has a calculator that estimates how many days a deposit needs to earn back its Base entry and exit costs.

What it does

• Tracks every active Base vault, up to 50. src/lib/beefy.ts keeps Beefy vaults on Base with status active, sorts them by TVL and takes the largest 50 (limit = 50). The cap is a ceiling, not a target. On 24 Oct 2026, Beefy's public /vaults endpoint listed 537 Base vaults: 26 active, 200 paused and 311 eol. The live /api/health reported 26 tracked vaults and 100% snapshot coverage the same day, so the count matches Beefy's active list. On 21 Oct 2026 /api/health had reported 50; I did not record Beefy's list that day, so I cannot say which vaults left. Beefy's /vaults API shows 200 Base vaults paused at once at 2026-10-21 23:00 UTC (the shared pausedAt 1792623600), and the monitor's daily export shows tracked vaults falling from 50 to 26 between the 22 Oct and 23 Oct scrapes, with summed TVL in the export falling from about $13.9M to about $3.7M (about $13.6–14.2M across 13–22 Oct, about $3.3–3.7M across 23–25 Oct). Each daily run marks vaults missing from the active list inactive (is_active = false).
• Compares realized and reported yield. For each interval, the expected return uses the APY recorded at the start of that interval, not today's APY. The method is in src/lib/drift.ts.
• Withholds results until the evidence is sufficient. A vault gets a drift result only after at least 3 valid observations that span at least 7 days. A vault goes to review when its drift is −5% or lower.
• Summarizes the portfolio. The dashboard shows capital analyzed, TVL-weighted expected and realized APY, an annualized yield gap (Σ TVL × (expected APY − realized APY)), and a recommended action.
• Estimates break-even days. The calculator at / can use the live headline APY, the time-weighted recorded APY, or realized share-price growth. It shows downside, current, and upside scenarios.
• Exposes its evidence. /api/export returns the valid observations from the rolling 90-day window as CSV or JSON. /api/health reports RPC, freshness, and coverage status. /api/summary returns the dated headline figures that the snapshot archive stores.

Headline figures, as of 25 Oct 2026 (UTC date of the latest daily snapshot, snapshots/2026-10-25.json):

Figure	Value
Capital analyzed	$3.17M (9 of 26 vaults pass the evidence gate)
Vaults requiring review	3
Annualized gap	$91.8K
Tracked vaults	26

These values move with every daily snapshot. The dated snapshot archive keeps each day's figures in the repository, so a number quoted anywhere can be traced to a date. Earlier figures differ: the 21 Oct 2026 capture showed an $89.5K gap on 50 tracked vaults, before Beefy's active list shrank to 26. That capture was not archived. The 25 Oct file was recomputed from the public /api/export rows with the dashboard code, and it matches the live dashboard ($3.2M, 11.52% expected, 8.62% realized, $91.8K, 3 vaults). Later files come from /api/summary on the live site.

How it works

1. Vercel Cron calls /api/cron/scrape daily at 01:15 UTC (vercel.json). The route requires the CRON_SECRET bearer token.
2. src/lib/beefy.ts loads vault, APY, TVL, and price data from the Beefy API.
3. src/lib/pps.ts reads getPricePerFullShare() and decimals() for every vault in one Multicall at a single Base block.
4. Ingestion functions in supabase/migrations/ write snapshots, failures, and the run status in one transaction. Unique keys on (chain, run_date) and (vault_id, snapshot_date) make a retry safe.
5. src/lib/drift.ts and src/lib/dashboard.ts compute interval-matched drift, the evidence gate, and the portfolio summary for src/app/dashboard/.

The project guide has the full architecture diagram, formulas, and database model.

Run it

Prerequisites: Node.js 20 or later (CI uses Node.js 20) and npm. The dashboard needs a Supabase project. The calculator works without one. The database tests need PostgreSQL 16 and psql.

git clone https://github.com/RahilBhavan/beefy-yield-fidelity-monitor.git
cd beefy-yield-fidelity-monitor
npm ci
npm run check    # lint, unit and route tests, production build

npm run check ran 53 Vitest tests on 25 Oct 2026: 52 passed and 1 skipped. The skipped test is the opt-in live integration test. The other suites:

npm run snapshot           # write snapshots/<as-of date>.json from the live /api/summary
npm run test:e2e           # 16 Playwright cases (8 tests × desktop and mobile Chromium), with axe WCAG A/AA checks
npm run test:integration   # live Supabase and Base RPC test; needs credentials and network access
npm run test:load          # smoke load on one endpoint (default /api/data); needs a running server

To run the app:

cp .env.example .env.local
npm run dev

Open the URL that npm run dev prints (Next.js uses port 3000 unless it is taken). Without Supabase settings, the dashboard shows an explicit "unconfigured" state. To collect data, set NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY, and CRON_SECRET in .env.local. Apply the files in supabase/migrations/ in filename order. Then start the first scrape, with APP_URL set to the URL npm run dev printed:

curl -H "Authorization: Bearer $CRON_SECRET" "$APP_URL/api/cron/scrape"

BASE_RPC_URLS is optional. The public Base RPC is the fallback. Keep the service-role key out of browser code.

CI (.github/workflows/ci.yml) runs npm run check, npm audit --omit=dev, every migration plus supabase/tests/production_fidelity_test.sql on PostgreSQL 16, and the Playwright suite.

Repository layout

• src/app/: pages (/, /dashboard, /dashboard/strategies, /methodology) and API routes (cron/scrape, data, export, gas-estimate, health).
• src/components/: calculator, portfolio table, and drift chart.
• src/lib/: Beefy adapter, PPS reads, gas model, drift method, and dashboard analysis, with their unit tests.
• supabase/migrations/: PostgreSQL schema, row-level security, and ingestion functions.
• supabase/tests/: database invariant tests.
• tests/e2e/: Playwright and axe browser tests.
• snapshots/: one JSON file per as-of date with the headline figures, plus index.json.
• scripts/: the snapshot writer and the load-smoke script.
• docs/: project documentation.
• .github/workflows/: CI and a production health probe that runs every 15 minutes.
• PRD.md: the product requirements document.

Documentation

• Project guide: architecture, formulas, database model, API routes, and verification.
• Case study: the decision problem, the analytical approach, and the recommendation framework.
• Operations runbook: release steps, service objectives, alerts, incident response, and rollback.
• Security model: credential boundaries, row-level security, and rate limits.
• Methodology: the public methodology page on the live site.

Sources

• Beefy API. vault list, APY, TVL, performance fee, and token prices. Each daily snapshot stores the APY and TVL from that run.
• Base chain. PPS and decimals read from each vault contract through Multicall. Each snapshot records the block number, block hash, and RPC provider label.
• Date range. Production collection started in August 2026 and adds one snapshot per vault per UTC day. The project does not backfill history.
• Freezing. The database keeps every snapshot. /api/export publishes the rolling 90-day window with its provenance, so you can recompute the dashboard figures.

Beefy data belongs to Beefy. The MIT license covers this project's code, not third-party data, names, or marks.

Scope and limits

• Independent portfolio project. Not affiliated with or endorsed by Beefy.
• Decision support, not financial advice. A flagged vault is a variance to investigate, not a conclusion about its cause.
• Base only. The project does not model token-price exposure, impermanent loss, slippage, bridge costs, taxes, exploit risk, or future APY.
• The gas estimate uses representative 500,000-gas deposit and withdrawal transactions. It is not a wallet-specific quote.
• Short windows make annualized figures volatile. Interval APY inputs above 1,000% stay in the export but are left out of the analysis.
• Live figures change daily. This README gives each headline figure with its as-of date and the archived file it comes from. Only 25 Oct 2026 is archived so far. The archive is written by running npm run snapshot; no scheduled job commits to it.

License

MIT. See LICENSE.

Part of Rahil Bhavan's projects.