feat(ethena): remove LlamaRisk backing check (API decommissioned)

monitoring.yaml

@@ -121,14 +121,12 @@ protocols:
     tasks:
       - protocols/ethena/ethena.py
     monitors:
-      - name: "USDe Backing Ratio (Ethena API)"
+      - name: "USDe Backing Ratio"
         description: "(collateral + reserve fund) / supply < 1.0 (CRITICAL) or < 1.005 (HIGH) via Ethena transparency API"
-      - name: "USDe Backing Ratio (LlamaRisk)"
-        description: "Independent (collateral + reserve fund) / supply < 1.0 (CRITICAL) or < 1.005 (HIGH) via LlamaRisk"
       - name: "On-Chain Supply Consistency"
-        description: "LlamaRisk USDe/sUSDe supply vs on-chain totalSupply() differs by > 0.5%"
+        description: "Ethena API USDe supply vs on-chain totalSupply() differs by > 0.5%"
       - name: "Data Freshness"
-        description: "Collateral, chain, or reserve data older than 12 hours"
+        description: "Supply or reserve fund data older than 3 hours"
 
   euler:
     display_name: "Euler"

protocols/ethena/README.md

@@ -6,35 +6,20 @@ Ethena is a synthetic dollar protocol built on Ethereum that provides a crypto-n
 
 ## Monitoring
 
-The script [`ethena/ethena.py`](ethena.py) runs daily via our VPS automation to sanity-check that **USDe remains fully backed** and that the public data feeds are fresh and internally consistent. Telegram messages are sent if some values are out of the expected range.
+The script [`ethena/ethena.py`](ethena.py) runs daily via our VPS automation to sanity-check that **USDe remains fully backed**. A Telegram message is sent if the backing ratio drops below the expected range.
 
-Two **independent** backing checks run each cycle — one against Ethena's own transparency API, one against LlamaRisk. They run in isolation (a failure or false positive in one provider never suppresses the other), and **every alert is prefixed with the provider that triggered it** (`[Ethena API]` or `[LlamaRisk]`) so it is obvious which source fired.
+The backing check (`ethena_backing_check`) computes `(collateral + reserve fund) / supply`, alerts **CRITICAL** if `< 1` and **HIGH** if `< 1.005` (`COLLATERAL_RATIO_TRIGGER`). USDe targets ~1:1 collateral backing with a separate reserve fund as the buffer, so the collateral-only figure hovers right around 1.0; including the reserve fund gives the true solvency ratio and avoids false positives on fractional collateral dips. Alerts are prefixed with `[Ethena API]` to make the source explicit.
 
-Both checks compute the same ratio: `(collateral + reserve fund) / supply`, alert **CRITICAL** if `< 1` and **HIGH** if `< 1.005` (`COLLATERAL_RATIO_TRIGGER`). USDe targets ~1:1 collateral backing with a separate reserve fund as the buffer, so the collateral-only figure hovers right around 1.0; including the reserve fund gives the true solvency ratio and avoids false positives on fractional collateral dips.
-
-> **The two checks measure different collateral lenses — they are not a like-for-like cross-check.** The Ethena check queries `/positions/current/collateral?latest=true`, whose `totalBackingAssetsInUsd` is a *net backing* figure that tracks supply ~1:1 (ratio ≈ 1.00, ≈ 1.015 with reserve). LlamaRisk (and the same Ethena endpoint *without* `latest=true`) report *gross collateral*, ~2.7% higher (ratio ≈ 1.027, ≈ 1.042 with reserve). They agree asset-by-asset to ~0.04%, so the ~2.7% ratio difference between the two checks is expected and definitional, **not** a data-staleness or backing problem. Each is a valid independent lower-bound on backing; they are intentionally kept separate rather than reconciled into one ratio.
-
-### Check 1 — Ethena Transparency API (`ethena_backing_check`)
+### Data Source — Ethena Transparency API
 
 Uses Ethena's own transparency API (`app.ethena.fi`). This API was previously blocked for GitHub Actions IPs, so a Chaos Labs / Oracle Security Proof-of-Reserve endpoint was used instead. That endpoint has since been decommissioned (returns HTTP 503), and Chainlink's USDe Proof of Reserves (Ethena's [PoR launch](https://ethena.fi/blog/usde-proof-of-reserves-launch) with Chainlink, Chaos Labs, LlamaRisk and Harris & Trotter) is not published as a public on-chain feed we can query. Since monitoring now runs on our VPS, Ethena's transparency API is reachable.
 
 1. **Supply**: `GET /api/solvency/token-supply?symbol=USDe`
 2. **Collateral**: `GET /api/positions/current/collateral?latest=true` (`totalBackingAssetsInUsd`)
 3. **Reserve fund**: `GET /api/solvency/reserve-fund` — latest point of the `queryIndex[0].yields` time series.
 
-### Check 2 — LlamaRisk (`llama_risk_check`)
-
-Uses the LlamaRisk transparency dashboard as a fully independent second opinion:
-
-`GET https://api.llamarisk.com/protocols/ethena/overview/all/?format=json`
-
-- **Backing ratio**: `(collateral_value + reserve_fund) / total_usde_supply`, same CRITICAL/HIGH thresholds as Check 1.
-- **On-chain cross-validation**: LlamaRisk's USDe and sUSDe supply are compared against on-chain `totalSupply()`; a MEDIUM alert fires if they differ by more than 0.5%. Skipped when LlamaRisk chain data is older than 2h (it would be out of sync with chain state).
-- **Data freshness**: LOW alerts if LlamaRisk collateral or reserve data is older than 12h.
-
-> NOTE: LlamaRisk data has historically lagged/diverged from Ethena's; it is treated as a secondary cross-check, which is why the two checks are independent and separately labelled rather than merged into one ratio.
+The API supply figure (the ratio's denominator) is additionally cross-checked against on-chain `totalSupply()` of the [USDe token](https://etherscan.io/address/0x4c9EDD5852cd905f086C759E8383e09bff1E68B3). A **MEDIUM** alert fires if the off-chain API and on-chain ground truth diverge by more than 0.5% (`SUPPLY_DIFF_TRIGGER`), catching a misreporting API.
 
-#### On-Chain feeds used by Check 2
+> **On the `latest=true` collateral figure:** it returns Ethena's *net backing* number, which tracks supply ~1:1 (ratio ≈ 1.00, ≈ 1.015 with reserve). The same endpoint *without* `latest=true` returns a detailed per-exchange breakdown whose total is *gross collateral*, ~2.7% higher — but that breakdown is a stale snapshot (items lag several hours). We use the fresh net figure plus the reserve fund as the buffer.
 
-1. **USDe Supply** — `totalSupply` for the [USDe token](https://etherscan.io/address/0x4c9EDD5852cd905f086C759E8383e09bff1E68B3)
-2. **sUSDe Supply** — `totalSupply` for the [sUSDe token](https://etherscan.io/address/0x9D39A5DE30e57443BfF2A8307A4256c8797A3497)
+> **Removed:** a second independent check against LlamaRisk's transparency API (`api.llamarisk.com/protocols/ethena/...`) previously ran alongside this one. LlamaRisk decommissioned that endpoint (now HTTP 404; the host only serves `aave-v4` routes), so the check was removed.