onvexia.com

GraphQL metrics endpoint — Metric/slug GraphQL dialect; point a compatible client's host at this URL. Supports getMetric{timeseriesData, timeseriesDataJson, timeseriesDataPerSlugJson, aggregatedTimeseriesData, metadata}, getAvailableMetrics, getAvailableSlugs, getMetricMetadata, allProjects, projectBySlug, getEcosystems. Whole-document cost budget: 500,000 estimated points per request.

AI substance — Does an AI-sector project actually ship code, against how much attention it gets. Four states: ships, silent, no_public_repo, and unmapped — the last means WE have not checked and is a gap in our coverage, never a finding about the project.

Webhook delivery log — Every alert notification attempt with its outcome. Failures carry a reason — if an alert never arrived, this says why. Requires authentication.

Get alert by ID — Retrieves a specific alert by ID.

List rule deliveries — Returns recent deliveries for one rule.

Get alert history — Retrieves trigger history for an alert.

Send test — Queues a signed test payload to every webhook destination on a rule.

Reveal secret — Returns the rule's signing key to its owner.

Rotate secret — Issues a new signing key, invalidating the old one.

Batch correlation — Returns correlation analysis for multiple assets.

Get correlation — Returns correlation analysis for an asset.

Get metric correlation — Returns correlation for a specific metric.

Get top creators — Returns top creators ranking.

Get creator rank — Returns a specific creator's rank.

Get leading indicators — Returns leading indicator analysis.

Get trading signals — Returns composite multi-pillar strength readings.

Get signals for asset — Returns composite strength readings for a specific asset.

Get active assets — Retrieves all active assets.

Get assets by category — Retrieves assets by category.

Add contract address — Adds a contract address to an asset.

Get asset by symbol — Retrieves an asset by its symbol.

Get the fundamental analysis brief — Market snapshot, TVL, revenue and derivatives positioning for one asset, composed into sections. Every section reports one of three states: `measured` (held, with each value naming its source and age), `not_held` (it exists for this asset and we do not have it, with the source and cost to get it), or -- in the separate `not_applicable` list -- a section this asset class cannot have, such as a treasury for a token that has no DAO. Revenue is split into the entity's OWN fees and, for a chain, the ECOSYSTEM total earned by protocols deployed on it; the two differ by ~112x for Arbitrum and only the first accrues to the token.

Token holder concentration — Top holders reconstructed from chain history, with entity labels where known. Every share is of the top 250 positions we store, NOT of total supply — supply is never read. Derived from Transfer events, so it misses rebasing tokens.

Get signal integrity — Alias of /v1/assets/{symbol}/signal-integrity. Is the hype on this token real or bot-driven?

Contract addresses per chain — 200 with an empty list for natively-issued coins like BTC — no contracts is a valid answer, not a 404.

Restatements touching one asset — Includes corpus-wide corrections, which affected this asset too.

Get asset scores — Retrieves Galaxy Score and AltRank for a specific asset.

Get score history — Retrieves historical scores for an asset.

Get signal integrity — Is the social attention on this token real or manufactured? Fuses bot probability, coordinated-campaign overlap, account age and engagement authenticity into a real-vs-fake score, with the components exposed so the number can be checked.

Social signal series — Per-asset social signal over time. `available` states plainly whether we hold it.

Get indicators and multi-timeframe support/resistance — Moving averages, Wilder RSI and ATR, fibonacci anchored to the measured impulse, and the volume profile -- reported per timeframe (1w/1d/4h/1h), plus support and resistance levels MERGED across those timeframes. `strength` is the sum over contributing timeframes of (timeframe weight x independent method families on that timeframe), so a level confirmed on four charts outranks one confirmed on the hourly alone; every method names the timeframe that produced it. An indicator is null only when the candle sample was too short, and `unavailable` says which and by how much. Also serves derived SETUPS -- a spot accumulation ladder and ranked long and short setups -- as geometry over those levels: an entry is a level cluster, a stop is that level offset by 0.6x the measured daily ATR, targets are the next opposing levels, and reward-to-risk is computed from those three prices rather than asserted. Nothing is forecast.

Get category rankings — Retrieves rankings within a category.

BigQuery spend ledger — What the on-chain backfills actually cost, by month. Every query is dry-run costed before it runs.

Exchange netflow — Daily exchange in/out flow with per-chain coverage stated beside it.

HODL waves — UTXO age-cohort snapshots per chain, with the maturity cutoff applied.

Per-chain coverage — Tokens mapped, labels held, whales seen, and the real attribution rate per chain.

Social coverage — Per-metric social coverage with its source and when each was computed — the unflattering numbers included.

Get creator rankings — Retrieves top creators.

Get creator rankings by platform — Retrieves top creators for a specific platform.

Emerging DEX pairs — New pairs above a liquidity floor, with the floor reported in the response.

Disclosure radar — Public-record documents about issuers we track — SEC filings and federal court dockets. A null severity means the document was found and NOT judged, which is not the same as benign; a name_unverified match may concern a different company.

Capital & attention flow — source -> asset -> narrative. Exchange venues from the 406 whale transfers carrying a named exchange (1.1% of the corpus, stated in the response); narratives from story headlines, because /v1/narratives labels are cluster ids rather than names.

Co-mention constellation — Asset co-mention graph: nodes, edges and what was filtered. Precomputed at write time — it measured 17.2 s before the rollup and 0.026 s after.

GraphQL metrics endpoint — Metric/slug GraphQL dialect for clients written against that shape. Also mounted at /graphql.

Get call-accuracy leaderboard — Influencers ranked by how often their price calls came true, by the Wilson lower bound of the hit rate so ten lucky calls cannot outrank a hundred good ones. Only creators with at least 20 resolved calls rank.

Get leaderboard — Crypto influencers ranked by MEASURED ACCURACY, not follower count. Each is scored Predictor or Reactor: did the call land before the move or after it. Ranked by a Wilson lower bound, so a lucky 11-call account does not outrank a steady 128-call one.

Get influencer call accuracy — Did this influencer's calls come true? Hit rate over 30, 90 and 180 days of calls at 24h, 7d and 30d horizons, with the Wilson lower bound, average return, excess over BTC and the predictor share. The headline is 90 days judged at 7 days.

Get influencer call receipts — Every price call an influencer made, as a receipt: the post excerpt and link, the asset, bullish or bearish, the entry price at the time of the call, and what happened after 24h, 7d and 30d — hit, miss or flat against a volatility-scaled threshold, with the return and the excess over holding BTC.

Get ledger — One influencer's full accountability record: every evaluated call, whether it preceded or followed the price move, hit rate and confidence-adjusted score.

Browse labelled addresses — Filter by chain, entity or category.

Identify an address — known:false means NO LABEL IS HELD. It does not mean the address is clean or unflagged.

Exchange listings — Recent exchange listing events.

Metric registry — Every metric with its parity and caveat. Unavailable metrics are included on purpose.

Batch metrics (many metrics × many assets) — One request, many series. Partial results are reported, never hidden: unavailable metrics and unknown assets come back with a reason instead of silently missing. Metered per SERIES, not per call. Caps: 25 metrics, 100 assets, 500 series, 250k points — exceeding one is a 400 showing the arithmetic, never a silent truncation.

Metric metadata — Parity, category, minimum interval and caveat.

Metric timeseries — Returns 404 naming the metric when unavailable. The caveat ships with the data for approximate metrics.

One metric across many assets — GET form of the batch endpoint — the shape a multi-line chart needs.

Narratives — LLM-clustered narratives across the corpus, with a note describing coverage.

Narrative rotation — Attention share per narrative and its movement against the equally long window immediately before, with the liquidity confirmation leg where an asset category exists. Narratives with no confirmation say so rather than reporting zero.

Analyze sentiment — Score any text for crypto sentiment, tuned for crypto slang and tickers. General models read 'going to zero' and 'wagmi' badly; returns polarity plus the terms that drove it.

Analyze sentiment batch — Score many texts for crypto sentiment in one call, same tuning as /v1/nlp/analyze.

Get aspect sentiment — Sentiment split by what people are actually discussing — technology, price, team and community. The aggregate can be flat while the parts disagree sharply.

Get entities — Entities extracted from posts about a token — projects, people, exchanges and protocols mentioned alongside it.

Get sentiment trends — How sentiment for a token has moved over time, so a spike can be told from a level.

What can be charted as candles — Per-asset candle coverage and its source venue. An asset absent here has no candles at all.

Candlestick datafeed — Real exchange klines (Binance, Bybit fallback). We do NOT synthesise candles from our price snapshots — those are ~1/day per asset, so the wicks would be invented. An asset with no listed pair returns available:false with a reason, never a flat series.

Pipeline health — Reports which tier cannot finish inside its own interval. WHY SATURATION IS THE HEADLINE AND NOT A LIST OF GREEN TICKS. Measured 2026-08-28: all nine hourly steps were SUCCEEDING, and the hourly tier was at 178.2% -- 6,416s of average work against a 3,600s interval. A status board would have been entirely green while the shared pipeline lock was held continuously, which is what left.

Asset reports — Generated per-asset research reports.

Asset report — The generated report for one asset.

When we restated history — Times our historical social metrics CHANGED because the matching rules were corrected. A number inside one of these periods may differ from what you saw before — that is a correction, not a market movement. An empty list means none were RECORDED, not that none happened.

Get top AltRanks — Retrieves top assets by AltRank.

Get top Galaxy Scores — Retrieves top assets by Galaxy Score.

Natural-language screen — Describe a screen in plain language and get the matching assets.

Asset screener — Filter the universe by field/operator. Call /v1/screener/fields first for the vocabulary.

Screener fields — Every field and operator the screener accepts, with usage — call before building a screen.

Analyze network — Analyse an arbitrary set of accounts for coordination: shared timing, shared phrasing and shared audiences.

Get coordinated campaigns — Coordinated shill and pump campaigns detected from posting-time clustering, phrase reuse and account-cohort overlap. Use before treating a trend as organic demand.

Share of social attention per asset

Get social graph — The reply and mention graph around one account — who amplifies whom, for separating a real conversation from one cluster talking to itself.

Get influencers — The accounts driving conversation about one token, ranked by measured reach and engagement in our corpus — not by follower count, and not by whether they were right.

Social posts — Raw collected posts, filterable by platform.

Similar posts — Embedding-nearest posts. Returns an explicit 404 naming the reason when the post has no embedding yet — embed_posts.py trails the newest documents — rather than an empty list.

Read-only SQL — Ad-hoc SELECT for questions the REST surface cannot express. Read-only, statement-timeout bounded, 10k-row cap, and truncation is always reported. See /v1/sql/schema for queryable relations.

What the SQL surface exposes — Relations and columns queryable via POST /v1/sql. User accounts, API keys, credit ledgers, alert rules and raw social posts are NOT in this schema and the query role has no permission to read them.

Stories — LLM-generated stories grouping related posts.

Posts behind a story — The source posts a story was assembled from — the receipts.

Resolve ticker — Answers "when people write TIA, do they mean the asset?".

Tokenized equity wrapper basis — One real-world underlying per row, priced by every issuer that tokenizes it (xStocks, Ondo, bStocks, Reality, Backpack, Robinhood, Figure, Securitize, Backed), with the spread between cheapest and dearest. Cross-wrapper only — comparing against the underlying stock needs a licensed equity feed.

Topic ranking — Ranked social topics for a window. Returns topics + window_hours.

Trending assets by social volume — Ranked by platform_trust-WEIGHTED mentions. Ships the corpus size behind the ranking and marks ties, because a bare rank is a stronger claim than a thin corpus supports. Measures ATTENTION, not sentiment or quality.

Search the asset universe — Resolve a loose reference into a real symbol. Ranks exact symbol, then prefix, then market cap.

Address watchlists — Watch an on-chain address; be told when it moves value. Every response carries the coverage bound — we index LARGE transfers only, so an empty feed means we observed nothing matching, NOT that the address was inactive.

What address watches can see — Per-chain indexed transfer counts and the observed USD floor, derived from the data rather than from configuration. Chains we accept but have not indexed are listed in no_data_yet.

Observed transfers for your watches — counterparty_label is null when WE HOLD NO LABEL — that is not a statement that the counterparty is unknown, and never that it is safe.

Get

Events — Returns one watch's observed transfers.

Get whale alerts — Recent large-transfer alerts across tracked assets and labelled addresses.

Subscribe whale alerts

Exchange accumulation vs distribution — Inflow to an exchange is distribution pressure; outflow is accumulation.

Get exchange flows — Net movement of a token into and out of exchange wallets — inflows are supply arriving where it can be sold, outflows are supply leaving to self-custody. A single large transfer is often an exchange rebalancing its own wallets, not a market action.

Get exchange flows summary — Aggregate exchange inflow and outflow across tracked assets, for spotting market-wide accumulation or distribution rather than a single token's move.

Get smart money movements — Wallets whose buying has historically preceded price strength — smart-money accumulation tracking across labelled addresses, with the flow direction and counterparty label.

Get top holders — Largest holders for an asset, reconstructed from chain history with entity labels where known. Shares are of the top positions stored, NOT of total supply.

Whale transfers — Recent large on-chain transfers.

Get whale transaction feed

Get wallet profile — Profile for a single wallet: entity label where known, holdings, and observed transfer behaviour. A null label means no label is held — never that the wallet is safe.

Get wallet transactions — Observed transfers for one wallet, with counterparty labels where known.