feat(ethena): run Ethena + LlamaRisk backing checks independently

monitoring.yaml

@@ -121,8 +121,12 @@ protocols:
     tasks:
       - protocols/ethena/ethena.py
     monitors:
-      - name: "USDe Backing Ratio"
-        description: "Collateral / supply < 1.0 via Ethena transparency API (CRITICAL)"
+      - name: "USDe Backing Ratio (Ethena API)"
+        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%"
       - name: "Data Freshness"
         description: "Collateral, chain, or reserve data older than 12 hours"
 

protocols/ethena/README.md

@@ -8,50 +8,33 @@ Ethena is a synthetic dollar protocol built on Ethereum that provides a crypto-n
 
 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.
 
-### Data Source - Ethena Transparency API
+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 primary backing check 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 and is used directly.
+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.
 
-1. **Supply**: `GET /api/solvency/token-supply?symbol=USDe`
-2. **Collateral**: `GET /api/positions/current/collateral?latest=true`
-3. **Backing Ratio**: `totalBackingAssetsInUsd / supply` — alert CRITICAL if `< 1`. USDe targets ~1:1 collateral backing with a separate reserve fund as the buffer, so the collateral-only ratio sits just above 1.0 in normal operation.
-
-### Data Sources - LlamaRisk
-
-> NOTE: LlamaRisk data is not reliable, so it is currently disabled (`llama_risk_check`).
+> **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.
 
-#### Off-Chain
+### Check 1 — Ethena Transparency API (`ethena_backing_check`)
 
-Data used is provided by Ethena on [transparency page](https://app.ethena.fi/dashboards/transparency) and LlamaRisk:
+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. **Ethena Transparency**
-   • Collateral: `GET /positions/current/collateral?latest=true`
-   • Supply  : `GET /solvency/token-supply?symbol=USDe`
-2. **LlamaRisk Dashboard**
-   `GET https://api.llamarisk.com/protocols/ethena/overview/all/?format=json`
-
-> NOTE: This LlamaRisk cross-check section is currently disabled (`llama_risk_check`). The note that Ethena data was unavailable applied to the old GitHub Actions setup; on our VPS the Ethena transparency API is reachable and is the primary source (see above).
+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.
 
-#### On-Chain
+### Check 2 — LlamaRisk (`llama_risk_check`)
 
-1. **USDe Supply**
-   `totalSupply` for USDe token
-2. **sUSDe Supply**
-   `totalSupply` for sUSDe token
+Uses the LlamaRisk transparency dashboard as a fully independent second opinion:
 
-#### What We Monitor
+`GET https://api.llamarisk.com/protocols/ethena/overview/all/?format=json`
 
-1. **Collateral Ratio**
-   `totalBackingAssetsInUsd + reserveFund / totalUsdeSupply`
-   • Warn if ratio < **1.01**
+- **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.
 
-2. **Dual-Source Consistency**
-   • Ethena vs LlamaRisk supply — alert if they differ by > 0.1%
-   • Ethena vs LlamaRisk collateral — alert if they differ by > 0.1%
+> 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.
 
-3. **Data Freshness**
-   • If collateral or chain data is older than 12h from either API triggers a stale-data warning. Also, if reserve data is older than 12 h, send a warning.
+#### On-Chain feeds used by Check 2
 
-4. **On-Chain Supply**
-   • Ethena vs LlamaRisk supply for USDe and sUSDe — alert if they differ by > 0.5%
-   • If chain data is old, use on-chain data for validating backings
+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)