Pricing Pipeline Changelog

Impact Notes

• `dexscreener-prices` now reflects only `/tokens/v1/{chainId}/{address}` availability in the late-stage stablecoin pricing fallback
• The symbol-search recovery path now records independently under `dexscreener-search`, which keeps search-specific failures visible without poisoning exact-address availability
• Public-health grouping excludes the search-only breaker so a best-effort addressless fallback issue does not count as a separate top-level availability circuit group

Impact Notes

• Pass 4 still prioritizes exact-target assets first and then larger circulating names, but it now walks the full sorted missing set until it spends the 10-request DexScreener budget
• Addressless non-unique symbols that are skipped without making a request no longer consume one of the candidate slots that can reach lower-rank unique-symbol recoveries such as CHFAU or ctUSD
• This reduces false `dexscreener-prices` breaker opens during bad network windows because the pass has more chances to record at least one healthy DexScreener response before giving up

Impact Notes

• Live post-enrichment validation now lets supported non-USD fiat assets replace materially divergent weak or mixed-source USD publications when a direct native quote implies a fresher peg-consistent USD mark
• The same native lane can now fill a missing live price for supported non-USD fiat assets when direct native CoinGecko pricing exists and the derived USD mark passes publication validation
• That native-implied mark remains a fresh fallback-validation lane rather than a replay-safe primary consensus source: it is not written into `price_cache`, and later replay cannot publish it as cached continuity

Impact Notes

• Historical CoinGecko market-chart replay now passes the configured CoinGecko API key through the backfill/admin path so native-fiat history can use the authenticated transport consistently during broad repairs
• Supported non-USD fiat backfills now replay native-fiat history at daily cadence with a two-point confirmation window across a 36-hour gap tolerance instead of opening on isolated thin hourly prints
• Extreme single-point native crashes of 5,000 bps or more are still preserved even when the normal historical confirmation rule would otherwise suppress the event

Impact Notes

• Supported non-USD fiat assets such as EUR, CHF, GBP, JPY, SGD, AUD, CAD, BRL, IDR, TRY, ZAR, PHP, MXN, RUB, and CNH/CNY can now consult fresh direct native-peg quotes before downstream depeg logic trusts a derived USD-versus-FX move
• Historical market replay now prefers direct CoinGecko native fiat pairs for those supported pegs and compares that native series directly to the `1.0` peg instead of replaying through `USD price / FX reference` when native history exists
• The published live USD price path still does not gain a second CoinGecko-derived consensus voice; this remains a downstream validation and historical-replay hardening change rather than a new cached live source

Impact Notes

• Live depeg detection now checks a fresh direct `coin/native-peg` quote for supported fiat pegs such as BRL before opening or extending downstream depeg state
• Pending depeg confirmation uses that same direct native quote first, reducing BRZ-style false positives caused by USD/FX reference drift
• The published live price remains the normal USD pipeline output; this change hardens downstream validation rather than introducing a new cached price source

Impact Notes

• Live pricing now treats `usdai-usd-ai` as a redeemable PYUSD wrapper and publishes it from the authoritative `protocol-redeem` lane when tracked PYUSD pricing is available
• Historical depeg backfills for `usdai-usd-ai` now replay the tracked PYUSD market series instead of trusting USDAI's own thin secondary-market history
• USDAI PegScore and depeg event history no longer inherit obvious false positives from wrapper-specific market dislocations that conflict with the token's instant-redemption semantics

Impact Notes

• Bunni slugs are now rejected during DEX crawl intake and DeFiLlama pool processing, so they no longer spend discovery or scoring budget
• Retained-pool filtering, challenger publication, and dex_prices publication all ignore Bunni even if stale or staged rows try to reintroduce it
• Pool-challenge replacement marks and promoted DEX bridge sources can no longer be pulled back toward peg by Bunni rows

Impact Notes

• USR no longer sits in `REDSTONE_TRACKED_SYMBOL_ALLOWLIST` once the live RedStone API stopped serving that exact-case symbol
• The RedStone price lane no longer spends request budget batching and retrying a symbol the provider currently omits
• Pricing-provider audit and merge-gate validation now fail only on live coverage drift that still exists, not on a known stale allowlist entry

Impact Notes

• dex_prices now rebuilds its aggregate and per-protocol bridge sources from retained priced pools after the full liquidity scoring filters run
• Promoted DEX bridge inputs and peg-summary dexPriceCheck now stay aligned with the same retained pool surface used by challenger publication and liquidity UI detail
• Skipped duplicate or low-quality discovery rows can no longer create synthetic near-peg DEX bridge marks for assets whose retained pools still show a depeg

Impact Notes

• Pool challenge still downgrades weak soft-source consensus when any qualifying challenger pool diverges beyond the peg-aware threshold
• Price replacement now requires at least two protocol-level challenger medians to diverge, rather than counting divergence from any one pool inside each protocol
• A rogue pool inside an otherwise agreeing protocol no longer drags severe depegs back toward peg on the published stablecoins snapshot
• The final replacement mark remains the TVL-weighted median across the corroborating divergent protocol groups

Impact Notes

• Pricing sources now carry explicit freshness kind, max trusted age, upstream-timestamp support, single-source authority, and market-capability metadata in the canonical registry
• High-confidence consensus now publishes the winning cluster median while preserving the internally selected cluster member for provenance and downstream policy
• DefiLlama list quotes now enter primary pricing as typed inputs with explicit observed-time provenance instead of inheriting mutable asset-state timestamps
• DEX bridge source freshness now preserves per-source timestamps from `price_sources_json` instead of flattening everything to the row write time
• DefiLlama contract fallback now prefers canonical tracked deployment identities, validates each quote before claiming the asset, and can probe multiple exact tracked coin ids when needed
• DexScreener fallback now prioritizes exact-target assets under the request cap and keeps symbol search reserved for addressless assets only
• Pyth confidence weighting is now smoother across medium-confidence prints, and RedStone requires at least 60% venue agreement before entering primary consensus
• The provider-config pricing audit is now part of CI validation instead of remaining a local-only check

Impact Notes

• DefiLlama pass 1 and pass 1b now validate quotes against the shared peg-aware bounds before marking an asset as resolved
• A bad DL contract response can no longer block later CoinMarketCap, Jupiter, or DexScreener fallback passes in the same run
• DexScreener symbol search now stays disabled whenever the asset already has a canonical chain+address lookup path; only addressless assets can use the unique-symbol search fallback
• This reduces wrong-identity recovery risk while preserving the exact-address DexScreener recovery path for assets with tracked deployments

Impact Notes

• GeckoTerminal probing now collapses pools to one TVL-weighted-median price per protocol before injecting a cross-protocol weighted-median mark
• Pool challenge replacement now uses corroborating divergent protocol groups rather than a raw all-pool weighted mean, making replacement marks less sensitive to repeated same-protocol pools
• Provider-config audit tests now guard CEX and RedStone coverage allowlists against duplicate or stale untracked entries
• The stale untracked `sUSDe` RedStone allowlist entry was removed so the runtime config matches the tracked registry again

Impact Notes

• Stablecoin payloads, peg-summary outputs, and `price_cache` rows now preserve `priceObservedAtMode` alongside `priceObservedAt`
• Primary consensus now carries freshness provenance per source and resolves a conservative effective mode for the selected price
• Hard single-source prices remain depeg-authoritative only when they retain source-native freshness provenance; local-fetch hard single-source prices now stay `confirm_required`
• Older rows remain backward-compatible: cached data that predates explicit freshness-mode storage does not automatically lose authority just because the mode is absent

Impact Notes

• Cached-fallback FX runs now keep probing Open Exchange Rates, Chainlink reference feeds, and gold-api.com instead of freezing their last known state until Frankfurter recovers
• A single stale intraday peg can no longer pin the whole FX lane in repeated cached fallback when OXR or Chainlink can refresh that subset independently
• If those independent probes restore fresh coverage for the expected fiat reference set, the run now exits cached fallback immediately and resets the fallback streak
• Operator metadata remains explicit about the failed Frankfurter / mirror path while no longer overstating the duration of an otherwise recovered FX incident

Impact Notes

• Previous-trusted price lookup now merges the last authoritative stablecoins publication with fresh replay-safe `price_cache` metadata
• Cached replay can keep publishing the last fresh corroborated depeg price through brief single-run corroboration gaps instead of dropping the asset to `N/A`
• Confirmed severe depegs no longer lose continuity just because an intervening stablecoins run published a `low` or unusable price state
• This closes the intermittent USR-style divergence where PSI could still explain the open depeg while detail surfaces lost the current price

Impact Notes

• Pricing source capabilities now come from one canonical registry shared by consensus, replay safety, pool challenge, GT probing, status health, and depeg trust classification
• Cached stablecoin payloads now preserve `priceObservedAt` and `priceSyncedAt`; compatibility `priceUpdatedAt` now reflects the true observation timestamp rather than the sync write time
• Soft single-source prices and soft-only high-confidence consensus can no longer mutate live depeg state directly; hard single-source sources such as Pyth, CEX, Curve, and protocol-redemption can still be authoritative
• Weak fixed-peg price jumps versus the previous trusted price now require corroboration before publication, closing the USR-style wrong-price path
• Pool challenge now uses the live $100K threshold in its published challenger snapshots and can harden weak soft-source outcomes, not only pre-downgrade high-confidence clusters
• GeckoTerminal probing now revisits weak CoinGecko / DL-list soft outcomes rather than only strict one-source cases
• Direct-API DEX quote conversion now reuses only authoritative tracked stablecoin prices; weak or stale tracked prices fall back to peg references instead of feeding the bridge loop
• Replay cache rows now keep source, confidence, observation time, sync time, and source lists; RedStone now derives its price from the venue median instead of the provider aggregate
• CoinMarketCap, Jupiter, and DexScreener enrichment passes now fail independently instead of aborting the whole late-enrichment block

Impact Notes

• Primary pricing candidates are no longer gated on `geckoId`; tracked assets can still enter consensus through Pyth, CEX, RedStone, Curve, DL-list, or DEX bridge inputs
• DefiLlama pass 1b now probes only tracked alternate deployments instead of synthesizing same-address identities across chains
• CoinMarketCap and DexScreener symbol fallbacks now require uniqueness within the tracked registry, reducing symbol-collision poisoning
• RedStone prices now require at least two corroborating venues before entering primary consensus
• Pool challenge now applies to DEX-inclusive soft consensus clusters unless an exempt hard source is present
• GeckoTerminal probing now cross-checks eligible single-source DL-list results in addition to single-source CoinGecko results
• Direct-API DEX pair conversion now prefers tracked cached stablecoin prices for quote legs and will not treat unknown addressed `USDC`/`USDT`-style symbols as automatic `$1` references
• Price-cache replay now stores only replay-safe non-low, non-fallback prices and the replay window is shortened from 24h to 6h
• Severe fixed-peg downside publication now requires corroboration unless the source is an explicit protocol-redemption or pool-challenge replacement mark

Impact Notes

• DEX identity is now canonical-only at runtime: DeFiLlama and subgraph parsing no longer mutate chain-aware token ownership
• Symbol fallback remains available only for addressless tokens; addressed unknown tokens are dropped instead of being reinterpreted by symbol
• DeFiLlama pools with `underlyingTokens` now match tracked assets by canonical addresses only, preventing positional symbol/address poisoning
• Promoted per-protocol DEX bridge sources are admitted into primary consensus only when corroborated or when no non-DEX voices exist
• The overlapping `dex-promoted` aggregate is withheld whenever promoted per-protocol DEX bridge data exists for the same asset

Impact Notes

• Quarter-hour FX runs no longer poison status simply because Frankfurter and mirror transports failed to re-deliver an already-current daily source snapshot
• Carry-forward runs preserve per-peg source dates and cadence metadata, so status still degrades normally once the underlying daily references actually age out
• Operator metadata still records the failed live transport path, but public health now aligns with source freshness rather than transport availability alone

Impact Notes

• Jupiter fallback now relies on V3 liquidity gates and peg-aware validation instead of treating optional `createdAt` metadata as a hard freshness cutoff
• Tracked Solana assets can recover through Jupiter even when V3 responses include old createdAt values alongside current block-level pricing
• DexScreener fallback now prefers exact chain+address pool lookups when an asset has a resolvable token address, reducing dependence on noisy symbol search results
• DexScreener search remains as the last fallback path, still capped by the shared request budget and liquidity sanity gates

Impact Notes

• Frankfurter remains the preferred ECB-backed business-day source for the core fiat set
• The existing `fawazahmed0/currency-api` mirrors still serve CNH/RUB/UAH/ARS and can backstop the wider fiat set when Frankfurter is unavailable
• When both current FX paths fail, ExchangeRate-API can now publish a daily full-set fiat snapshot instead of forcing an immediate cached-fallback run
• The About page and pricing methodology now disclose ExchangeRate-API as an externally visible FX reference source

Impact Notes

• CNH/RUB/UAH/ARS still use the secondary daily feed as their normal source path
• When Frankfurter fails, the fresher `fawazahmed0/currency-api` mirror can now populate the broader fiat FX set instead of forcing an immediate cached-fallback run
• Per-peg FX metadata preserves calendar-daily cadence and source-date semantics during this live fallback path
• Public health no longer needs to report long consecutive cached-fallback FX runs for a Frankfurter-only outage when the secondary feed is healthy

Impact Notes

• Pool challenge no longer depends on dex_liquidity.top_pools_json, so display truncation cannot hide a large challenger pool
• Published challenger snapshots are coverage-gated per stablecoin and fall back safely during migration gaps
• Cached FX fallback runs preserve per-peg source timestamps and source modes instead of refreshing them implicitly
• Health and status now report usable FX freshness, underlying source freshness, and consecutive fallback runs separately
• Non-USD and commodity validation consumers now read shared FX state instead of inferring freshness from cache updated_at alone

Impact Notes

• Kraken joins primary consensus at weight 2 for supported USD pairs
• Bitstamp joins primary consensus at weight 1 as a lower-weight corroborating CEX venue
• Primary CEX fetches remain grouped so the quarter-hour pricing lane does not add new peak connection fan-out
• Missing Solana prices can now resolve through Jupiter before DexScreener, gated by liquidity and peg-aware plausibility checks
• Curated Chainlink EUR/USD, GBP/USD, JPY/USD, XAU/USD, and XAG/USD feeds can now refresh the FX/reference cache when fresh and aligned
• Status source distribution now reports Kraken, Bitstamp, and Jupiter participation explicitly

Impact Notes

• Transitive source chains can no longer create fake multi-source high confidence
• Equal-size cluster ties now resolve deterministically by weight, spread, peg proximity, then label
• Fixed-peg assets no longer silently fall back into NAV-style 500 bps clustering when peg references are missing
• Stale or aggregate-only RedStone entries are excluded before consensus
• Protocol-backed redemption prices can no longer be overwritten by the GeckoTerminal probe
• Direct-API pools must pass shared TVL sanity gates before they suppress overlapping DeFiLlama pools

Impact Notes

• Fluid, Balancer, Raydium, and Orca now contribute at most one promoted consensus source each
• Repeated high-TVL pools from the same protocol can no longer overweight primary consensus by appearing multiple times
• dex_prices.price_sources_json now stores per-protocol aggregates for the pricing bridge
• Pool challenge no longer depends on dex_prices.price_sources_json; it reads current top pools from dex_liquidity instead
• Non-USD tracked stablecoin pairs use peg-reference-aware conversion when deriving direct-API DEX prices

Impact Notes

• Pool-level individual prices added as fourth depeg confirmation source — fixes dUSD-like depegs going undetected
• Pool challenge threshold now peg-type-aware: 300 bps for non-USD (was 500 bps for all)
• Pyth feeds with >200 bps confidence excluded from consensus; 100-200 bps downweighted
• RedStone excluded when internal venue agreement < 50%
• CG+DL-only consensus downgraded from high to single-source (illusory agreement)
• NAV tokens (FPI) now visible in peg-summary API with null deviation
• Full source list preserved in consensus label (no more truncation)
• Protocol override divergence warnings logged when >100 bps from consensus

Impact Notes

• Dropped DL coins API from primary consensus — it returned CG-sourced data, making CG+DL agreement tautological
• Added DefiLlama stablecoins list price (weight 1) as a genuinely independent aggregator voice
• Added GeckoTerminal pool probe (weight 1) for single-source CG-only assets with $10K TVL gate
• Pool challenge guard: downgrades soft-only high confidence to 'low' when any $100K+ TVL DEX pool diverges ≥500 bps
• Pool challenge price correction: replaces soft consensus price with TVL-weighted mean of all qualifying individual pool prices
• DEWS scoring suppresses degradation bonus for high→single-source transitions to prevent alert spikes
• ~130 assets retain genuine high confidence via CG+DL-list agreement; ~27 gain GT cross-check

Impact Notes

• 8 independent price sources with per-source circuit breakers and configurable weights
• Consensus algorithm clusters sources within 50 bps, picks highest-weight in largest cluster
• Authoritative protocol-redemption overrides for wrapper assets (cUSD, iUSD, crvUSD)
• 4-pass enrichment pipeline for assets still missing prices after primary consensus
• Price confidence tagging: high (2+ agree), single-source, low (disagree), fallback
• CoinMarketCap enrichment optimized from per-slug to batch listings endpoint

Impact Notes

• CoinGecko primary prices with DefiLlama cross-validation
• Basic price reasonableness checks against peg references
• DexScreener enrichment for assets missing from aggregators