BRS Signals API

Get the latest VAO value with all components.

Get historical VAO values.

Generate a trade signal with risk management.

Quick health check.

Full system status with all component health.

Run a backtest with the current configuration.

Get decoder signal statistics for dashboard indicator.

Get decoder signal statistics for dashboard indicator.

Pipeline funnel: how many cycles died at each gate. Public diagnostic — read-only counters from runtime_kv. Answers "why no signals?" in one glance. Query one of: (nothing) today ?day=YYYY-MM-DD a specific UTC day ?days=30 the last N UTC days, summed ?since=launch every day on record, summed ?since=YYYY-MM-DD from that day onward, summed The response includes `gate_order` — the decoder's evaluation sequence. Clients draw a survival funnel by subtracting each gate's count in that order; without it they have to guess, and a guess drifts the moment a gate is added.

Per-engine input stream coverage — what each eye is actually seeing. Q12: reads the SHARED DB-backed health store (vao.health_store), the union of evidence from both the API and live processes, so an API restart no longer zeroes the display and live-only streams (prediction_markets, whale_tracker, the engines) report truthfully. Every stream shows real evidence: status ok/stale/never, the actual last-seen timestamp (not "now"), age, freshness window, and which process produced it. The old hardcodes (`atr: True`, `orderbook_churn` defaulting True) are gone — health is derived from beats, never declared. `missing` names non-ok streams so nobody has to diff 5/6 against a mental list. Morning ritual: check `live_process` first, then coverage. Any eye below 50% → read `missing` and each entry's `reason`.

Get the current trade signal from the directional decoder (DB-backed, survives restarts).

Get the current trade signal from the directional decoder (DB-backed, survives restarts).

Per-call x402 access to the Pro bias endpoint (K3 wiring flags). Flow: no key or no proof → 402 (per-call $0.01) · agent pays → retries with tx_signature → VERIFY settlement FIRST (settle-after-charge) → charge the per-call ledger → serve the bias payload for this query. A verified-but-unrecorded settlement (contention) is REFUSED — never serve a paid response with no ledger line. A Pro key just serves (no per-call). The verify price is SERVER-controlled — the caller cannot name an `expected_amount` (K3 pin, Aug 26).

Premium tier: recent decoder decisions. Use ?active=true for non-suppressed BUY/SELL only.

Premium tier: recent decoder decisions. Use ?active=true for non-suppressed BUY/SELL only.

Live data counters (keyless). Every number here is read from the DB — nothing is hard-coded and nothing is inferred: - `signals_sent` — the authoritative all-time count from `decoder_decision_records`, the SAME query the funnel (`/api/v2/health/funnel`) and the landing gauntlet use, so both surfaces report one number. (`signals_fired` is retained as a back-compat alias of the same value; it must never be sourced from the legacy, empty `trade_signal_records`.) - `data_points` — rows actually stored in the fee-histogram and funding-divergence streams. - `days_collecting` — the true span of those rows, `MAX(timestamp) - MIN(timestamp) + 1` in days. It used to be the literal `14`; the real collection window is far longer, so it is now derived. 0 when there is no data yet (absence is not 14).

Bundled dashboard endpoint — regime meta + signal + suppressed + performance.

Regime transitions within the price window — for colored chart line segments.

Regime transitions within the price window — for colored chart line segments.

Get BTC closes for charting. Uses daily candles for >41 days (Binance 1000-candle limit).

Public track record — what our collectors captured, what BTC did next. Data-first design: 1. data_events: regime events + fee histograms + funding divergence (FREE tier) 2. decoded_signals: directional decoder decisions (PREMIUM tier) Never framed as "we were right" — just transparent data for quants.

Public track record — what our collectors captured, what BTC did next. Data-first design: 1. data_events: regime events + fee histograms + funding divergence (FREE tier) 2. decoded_signals: directional decoder decisions (PREMIUM tier) Never framed as "we were right" — just transparent data for quants.

Latest X-Ray sensor verdict.

Latest Pulse sensor verdict.

Latest Shadow sensor verdict.

Current convergence score + entropy gradient + all three engine verdicts.

Current convergence score + entropy gradient + all three engine verdicts.

Daily BTC-perpetual liquidation context for one UTC day. Retrospective daily context for the ORE archive — the reported USD notional of forcibly closed long/short BTC perpetual positions across covered venues. It is never a count of people or a measure of individual losses.

Compact per-day liquidation totals for the whole archive. One entry per UTC day (long/short/total USD + coverage), sorted by date. The ORE archive uses this to mark the most heavily liquidated days.

Current dealer gamma exposure + flip level.

Current dealer gamma exposure + flip level.

Signal performance grouped by regime × zone. Pure data — no win-rate claims.

Signal performance grouped by regime × zone. Pure data — no win-rate claims.

Return the public track record as a documented, versioned proof schema. Every record carries a signal id, emitted timestamp, side, confidence, entry price + price source, and fixed +4h/+24h outcomes (with best/worst excursion and resolution status). The envelope carries schema_version, methodology_version and a sample-size + Wilson-uncertainty summary so an agent can compute honesty metrics directly. Never includes suppressed or WAIT rows.

Canonical CSV export of the proof schema (same fields, stable order).

Server-rendered proof page — no JS required to see the evidence.

The signed proof manifest — the tamper-evident record behind the proof. Keyless, like the rest of the proof surface. Returns the seal (chain head, record count, signing public key, Ed25519 signature) and every chained record. Verification is offline: the public key and head hash are inside the seal, so a reader needs nothing but this payload to detect a post-hoc edit (a record hash breaks) or truncation (the seal no longer matches).

Generate a free API key.

Email-free machine key issuance (Design B, Aug 31). Mints a FREE key with no human in the loop. `owner_label` names the identity (NOT an email); `?ref=` records registry attribution. There is no email, so there is no email-based reissue path — holding the key IS the identity. Rate-limited per client IP so a script cannot mass-mint.

Return the key provisioned for a completed Checkout Session. The key is created server-side inside the webhook, so the customer's browser has no way to learn it after the redirect — which is why they were landing on "no API key" seconds after paying. Guarded by the session id itself: unguessable, and known only to whoever completed that checkout. We re-confirm with Stripe that the session is actually paid before returning anything. The redirect usually beats the webhook, so 202 means "come back in a moment" and the page polls.

Create a Stripe Checkout session for an arbitrary support amount. Used by ORE·Signals' download pages ("support — pay what you want", default $20). Grants nothing — there is no key or tier — so the webhook recognises the session by its ``kind=donation`` metadata and skips provisioning. CORS for ore-signals.com is already allowed in server.py.

The record page. Its freshness comes from the JSON it fetches, not from re-reading the shell — no-cache on the HTML is enough.

Serve the Central Command admin dashboard.

Serve the checkbox calendar planning board (data is admin-gated).

Serve the Wyckoff visual-layer study surface (data is admin-gated).

Serve the Freqtrade integration guide (rec 9 — surfaced on every plan).

Serve the usage / case study page.

Serve the legal/disclaimer page.

Serve the pricing page (x402-list site pillar).

Serve the terms-of-service page (x402-list site pillar).

Serve the favicon at the root path.

Landing page. (A 'gate middleware' is referenced in old comments; none is registered — the page is public.)