Indexer oracle /api/v1/oracle/price returns LUNC/USD exactly 100x too high

leonardocolucci commented

2026-08-10 16:08:24 +00:00

(Migrated from gitlab.com)

Summary

/api/v1/oracle/price on the production indexer returns LUNC/USD scaled exactly 100× too high, while attributing the values to KuCoin/MEXC — whose real same-minute prices are 100× lower.

Evidence (fetched within the same minute, 2026-08-10 ~16:06 UTC)

Source	LUNC/USD
indexer oracle (`price_usd`)	0.005039
— oracle's claimed "kucoin" source	0.00502
— oracle's claimed "mexc" source	0.00511 (earlier fetch)
KuCoin actual (`/api/v1/market/orderbook/level1?symbol=LUNC-USDT`)	0.00005009
MEXC actual (`/api/v3/ticker/price?symbol=LUNCUSDT`)	0.00005024
Binance actual	0.00005024
CoinGecko	0.00004996

The oracle's per-source values are precisely the real exchange prices × 100, so this looks like a unit/scaling bug in the fetcher (e.g. parsing a 1000LUNC or cents-denominated field, or a stray ×100), not a stale price.

Impact

Anything consuming this endpoint — dApp USD displays, charts, integrators valuing cLUNC inventory — overstates USD values by 100×. We caught it because our bot's portfolio meter suddenly claimed a ~$380 wallet was worth ~$38,000; a treasury or listing decision made off this feed would be off by the same factor.

Happy to retest once fixed.

## Summary `/api/v1/oracle/price` on the production indexer returns LUNC/USD scaled **exactly 100× too high**, while attributing the values to KuCoin/MEXC — whose real same-minute prices are 100× lower. ## Evidence (fetched within the same minute, 2026-08-10 ~16:06 UTC) | Source | LUNC/USD | |---|---| | indexer oracle (`price_usd`) | **0.005039** | | — oracle's claimed "kucoin" source | 0.00502 | | — oracle's claimed "mexc" source | 0.00511 (earlier fetch) | | KuCoin actual (`/api/v1/market/orderbook/level1?symbol=LUNC-USDT`) | 0.00005009 | | MEXC actual (`/api/v3/ticker/price?symbol=LUNCUSDT`) | 0.00005024 | | Binance actual | 0.00005024 | | CoinGecko | 0.00004996 | The oracle's per-source values are precisely the real exchange prices × 100, so this looks like a unit/scaling bug in the fetcher (e.g. parsing a 1000LUNC or cents-denominated field, or a stray ×100), not a stale price. ## Impact Anything consuming this endpoint — dApp USD displays, charts, integrators valuing cLUNC inventory — overstates USD values by 100×. We caught it because our bot's portfolio meter suddenly claimed a ~$380 wallet was worth ~$38,000; a treasury or listing decision made off this feed would be off by the same factor. Happy to retest once fixed.

PlasticDigits commented

2026-08-11 00:36:26 +00:00

(Migrated from gitlab.com)

Approved to fix. Likely cause is api confusing - the price api should be delivering ustc price. Breaking change to v1 approved (as we are still in non-economic phase) to change the api routes to /api/v1/oracle/price/ustc and /api/v1/oracle/price/lunc, with /api/v1/oracle/price listing available tickers such as {metadata: "... explain how to use the api here...", tickers:["ustc","lunc"]

PlasticDigits commented

2026-08-11 00:50:30 +00:00

(Migrated from gitlab.com)

mentioned in commit aa59731ea0

mentioned in commit aa59731ea0848b6d00c3392a2dcf7da8ed22a694

PlasticDigits commented

2026-08-11 00:50:43 +00:00

(Migrated from gitlab.com)

mentioned in merge request !1050

PlasticDigits commented

2026-08-11 00:50:50 +00:00

(Migrated from gitlab.com)

Fix implemented in !1050 (fix/515-oracle-price-tickers).

Breaking v1 change (as approved):

GET /api/v1/oracle/price → catalog { metadata, tickers: ["ustc","lunc"] }
GET /api/v1/oracle/price/ustc / /lunc → ticker snapshots
Matching history catalog + /history/{ticker}
Regression: make verify-issue-515

Please retest LUNC path against KuCoin/MEXC after indexer deploy.

Fix implemented in !1050 (`fix/515-oracle-price-tickers`). Breaking v1 change (as approved): - `GET /api/v1/oracle/price` → catalog `{ metadata, tickers: ["ustc","lunc"] }` - `GET /api/v1/oracle/price/ustc` / `/lunc` → ticker snapshots - Matching history catalog + `/history/{ticker}` - Regression: `make verify-issue-515` Please retest LUNC path against KuCoin/MEXC after indexer deploy.

PlasticDigits (Migrated from gitlab.com) closed this issue

2026-08-11 01:07:07 +00:00

PlasticDigits commented

2026-08-11 01:07:08 +00:00

(Migrated from gitlab.com)

mentioned in commit 92d5e3a105

mentioned in commit 92d5e3a1053c029fe04c64304761d90435da0e66

PlasticDigits commented

2026-08-11 01:07:21 +00:00

(Migrated from gitlab.com)

Merge sanity check (!1050 → main)

Merged via glab mr merge 332 after resolving a post-!1049 conflict in Makefile .PHONY (kept both verify-issue-512 and verify-issue-515). No automerge; CI was ci_quota_exceeded, not waited on.

Local checks on the MR branch: cargo test --lib oracle (11/11), cargo test --test api_oracle (10/10), Protocol/Trade frontend tests (33) — all passed. make verify-issue-515 needs frontend node_modules on a fresh checkout.

Acceptance vs #515

Criterion	Result
Bare `/api/v1/oracle/price` + `/history` → ticker catalog	Met
Snapshots at `/price/{ustc\|lunc}` and `/history/{ticker}`	Met
Poller fetches both tickers with distinct CEX/CoinGecko symbols	Met
Volume USD / overview still USTC-only	Met
Protocol page default `getOraclePrice()` → USTC/USD	Met
Docs/invariants X1–X6 + verify target	Met

Remaining gaps / follow-ups (ops, not code blockers)

Deploy checklist — Run migration 20260811000000_oracle_prices_multi_ticker.sql (indexer auto-migrate on start), deploy indexer + frontend together, and notify integrators: bare GET /api/v1/oracle/price is now a catalog, not a numeric price (use /price/ustc or /price/lunc).
verify-issue-515 bootstrap — Script bootstraps Postgres but not frontend deps; fresh checkouts fail the frontend step with vitest: not found until npm install in frontend-dapp.
Live CEX smoke — Symbol wiring is unit-tested; compare /price/lunc vs live KuCoin/MEXC LUNCUSDT after deploy (MR test plan).

Please retest LUNC vs USTC scale on staging/prod after indexer deploy.

## Merge sanity check (!1050 → main) Merged via `glab mr merge 332` after resolving a post-!1049 conflict in `Makefile` `.PHONY` (kept both `verify-issue-512` and `verify-issue-515`). No automerge; CI was `ci_quota_exceeded`, not waited on. Local checks on the MR branch: `cargo test --lib oracle` (11/11), `cargo test --test api_oracle` (10/10), Protocol/Trade frontend tests (33) — all passed. `make verify-issue-515` needs frontend `node_modules` on a fresh checkout. ### Acceptance vs #515 | Criterion | Result | |---|---| | Bare `/api/v1/oracle/price` + `/history` → ticker catalog | Met | | Snapshots at `/price/{ustc\|lunc}` and `/history/{ticker}` | Met | | Poller fetches both tickers with distinct CEX/CoinGecko symbols | Met | | Volume USD / overview still USTC-only | Met | | Protocol page default `getOraclePrice()` → USTC/USD | Met | | Docs/invariants X1–X6 + verify target | Met | ### Remaining gaps / follow-ups (ops, not code blockers) 1. **Deploy checklist** — Run migration `20260811000000_oracle_prices_multi_ticker.sql` (indexer auto-migrate on start), deploy indexer + frontend together, and notify integrators: bare `GET /api/v1/oracle/price` is now a **catalog**, not a numeric price (use `/price/ustc` or `/price/lunc`). 2. **`verify-issue-515` bootstrap** — Script bootstraps Postgres but not frontend deps; fresh checkouts fail the frontend step with `vitest: not found` until `npm install` in `frontend-dapp`. 3. **Live CEX smoke** — Symbol wiring is unit-tested; compare `/price/lunc` vs live KuCoin/MEXC LUNCUSDT after deploy (MR test plan). Please retest LUNC vs USTC scale on staging/prod after indexer deploy.

leonardocolucci commented

2026-08-11 09:30:24 +00:00

(Migrated from gitlab.com)

Retested — confirmed fixed, and thanks for the fast turnaround (filed → fixed in under a day).

Verification, same-minute against external references:

Ticker	Oracle	External (KuCoin)
`/api/v1/oracle/price/lunc`	0.00005005	0.00004982	✓
`/api/v1/oracle/price/ustc`	0.0050235	0.00503	✓

Also happy to correct my own diagnosis: this was never a 100× scaling bug — the old unlabeled endpoint was serving USTC/USD, and consumers (us included) read it as LUNC/USD. The "exactly 100×" in my title was just the coincidental USTC:LUNC price ratio. Your fix — ticker in the path, both assets served explicitly — is the right structural cure, and having the bare /price endpoint return metadata instead of a number is a nice touch: any stale consumer breaks loudly instead of silently misreading forever.

We've migrated our tooling to /price/lunc and keep independent exchange cross-checks as standard practice. Fine to close from our side.

Retested — **confirmed fixed**, and thanks for the fast turnaround (filed → fixed in under a day). Verification, same-minute against external references: | Ticker | Oracle | External (KuCoin) | | |---|---|---|---| | `/api/v1/oracle/price/lunc` | 0.00005005 | 0.00004982 | ✓ | | `/api/v1/oracle/price/ustc` | 0.0050235 | 0.00503 | ✓ | Also happy to correct my own diagnosis: this was never a 100× *scaling* bug — the old unlabeled endpoint was serving USTC/USD, and consumers (us included) read it as LUNC/USD. The "exactly 100×" in my title was just the coincidental USTC:LUNC price ratio. Your fix — ticker in the path, both assets served explicitly — is the right structural cure, and having the bare `/price` endpoint return metadata instead of a number is a nice touch: any stale consumer breaks loudly instead of silently misreading forever. We've migrated our tooling to `/price/lunc` and keep independent exchange cross-checks as standard practice. Fine to close from our side.

PlasticDigits commented

2026-08-15 09:36:29 +00:00

(Migrated from gitlab.com)

marked as related to #520