API

Every endpoint sits under one versioned base URL and returns JSON. Response fields are camelCase throughout — expectedMove, not expected_move — including inside nested objects. Query parameters stay literal and lower-case: symbol, fh, days, slug.

GET https://marketmania.ai/api/v1/indices/outlook/?symbol=BTC&fh=4h
→ 200 application/json

Every timestamp inside a payload — generatedAt, ts, latestTs — is ISO-8601 in UTC, for example 2026-08-04T09:01:00+00:00. Dates written for people, in this reference and across the site, are rendered en-GB instead, for example 4 Aug 2026: UTC for machines, en-GB for readers.

The indices are public by design. There is no key, no header and no account needed to read any endpoint on this page: they are anonymous, identical-for-everyone reads of a table a scheduled job has already materialised. A 401 from one of these paths would be a bug, not a state to handle.

Responses are page-cached for 60 seconds, so polling faster than that returns the same bytes. Requests share the public benchmarks rate limit; exceeding it returns 429. The indices are recomputed on a schedule, not per request — the freshest data any caller can get is the latest slot, which each payload states in generatedAt and latestTs.

symbol is one of ALL, BTC, ETH, SOL, BNB, XRP and defaults to ALL. fh is one of 1h, 4h, 1d, 1w, 1m and defaults to 4h. Both vocabularies are also published by the catalog endpoint, which is the safe thing to build a client against.

{
  "generatedAt": "2026-08-04T10:09:42.165465+00:00",
  "variant": "v0",
  "symbols": ["ALL", "BTC", "ETH", "SOL", "BNB", "XRP"],
  "horizons": ["1h", "4h", "1d", "1w", "1m"],
  "indices": [
    { "slug": "outlook",     "description": "…", "latestTs": "2026-08-04T09:01:00+00:00" },
    { "slug": "calibration", "description": "…", "latestTs": "2026-08-03T00:00:00+00:00" },
    { "slug": "smart",       "description": "…", "latestTs": "2026-08-04T09:01:00+00:00" },
    { "slug": "smart_daily", "description": "…", "latestTs": "2026-08-03T00:00:00+00:00" }
  ]
}

A latestTs of null means the index has published no row at all yet. The two daily slugs carry a UTC midnight, not a live slot: they are rollups of a closed day, so the newest complete row is normally yesterday's.

GET /api/v1/indices/outlook/?symbol=ALL&fh=4h

{
  "generatedAt": "2026-08-04T10:09:39.929001+00:00",
  "symbol": "ALL",
  "fh": "4h",
  "horizons": [ { …row per horizon… } ],
  "scorecard": [
    { "fh": "1h", "n": 465, "hitPct": 49.2473, "strongN": 358, "strongHitPct": 49.7207 }
  ],
  "acrossInstruments": [ { …row per instrument at fh… } ]
}

Arrays above are abridged. horizons holds one row per horizon for symbol; acrossInstruments holds one row per instrument at fh, ALL first. Each item carries its own ts, because a quiet coin's latest row can be older than its neighbours'.

GET /api/v1/indices/outlook/history/?symbol=BTC&fh=4h&days=30

{ "symbol": "BTC", "fh": "4h", "days": 30, "n": 178,
  "points": [ { "ts": "…", "sentiment": -3.3869, "label": "Sideways" } ] }

GET /api/v1/indices/calibration/?symbol=ALL

{ "symbol": "ALL", "ts": "2026-08-03T00:00:00+00:00", "date": "2026-08-03",
  "gap": -16.8013, "stated": 63.8365, "actual": 47.0352, "n": 10962,
  "buckets": [ { "bucket": "50-60", "lo": 50, "hi": 60, "n": 2262,
                 "stated": 56.9089, "actual": 47.7454, "gap": -9.1636 } ],
  "byModel":   [ { "model": "claude-fable-5", "n": 1467, "stated": 60.4901,
                   "actual": 46.9666, "gap": -13.5235, "discrimination": -2.045 } ],
  "byHorizon": [ { "fh": "1h", "n": 6682, "stated": 63.7206,
                   "actual": 48.683, "gap": -15.0376 } ],
  "below50":   { "n": 0, "stated": null, "actual": null } }

GET /api/v1/indices/smart-consensus/?symbol=ALL&fh=4h

{ "generatedAt": "…", "symbol": "ALL", "fh": "4h",
  "latest": { "ts": "2026-08-04T08:01:00+00:00", "score": -2.8571, "label": "Bearish",
              "majorityScore": -2.8571, "nForecasts": 70, "nModels": 7,
              "resolved": false, "realizedMovePct": null, "realizedDir": null,
              "weights": [ { "model": "claude-fable-5", "weight": 0.1429 } ] },
  "weights":  [ { "model": "claude-fable-5", "weight": 0.1429 } ],
  "accuracy": [ { "date": "2026-07-17", "ts": "…", "smartHitPct": 50.0,
                  "majorityHitPct": 50.0,
                  "bestModel": { "model": "qwen-3.7-max", "hitPct": 63.8298, "n": 47 },
                  "n": 12 } ],
  "scorecard": [ { "fh": "4h", "n": 115, "hitPct": 47.8261,
                   "strongN": 91, "strongHitPct": 46.1538 } ] }

GET /api/v1/indices/outlook/?fh=2h
→ 400

{ "errors": { "fh": ["unknown horizon '2h', must be one of ['1h', '4h', '1d', '1w', '1m']"] },
  "error": "fh: unknown horizon '2h', must be one of ['1h', '4h', '1d', '1w', '1m']" }

Both keys are always emitted: errors is the field-keyed shape the rest of the site reads, error is the flat one-liner. Neither consumer needs a special case.

The following are planned and are listed here so that their absence is explicit rather than implied. None of them exists on /api/v1/indices/ today, and nothing above depends on them.