Proof in the open. Picks with the desk.
Two public JSON endpoints under https://goseer.ai/api/ let anyone verify the engine is alive and that historical predictions calibrated. The four endpoints carrying actual forward picks (convergence, FOMC posteriors, predmarkets actionables, Trump claims) sit behind the same /desk auth that gates the trading view. Partners ping hello@goseer.ai for credentials.
All endpoints reflect the most recent engine run — finished 9h ago · /status for the full pipeline picture.
Quickstart — public endpoints
# Latest engine run + per-pipeline status
curl https://goseer.ai/api/status.json | jq '.ok, (.pipelines | length)'
# Cross-engine direction accuracy across resolved walk-forward predictions
curl https://goseer.ai/api/receipts.json | jq '.summary'
Partner endpoints require --user $SEER_USER:$SEER_PASS on the curl (HTTP basic). Same credentials as /desk.
Public endpoints
no auth · proof of calibrationLive orchestrator status. Most recent run's pass/fail per pipeline, durations, timestamps. Same data the homepage 'Engine: N/M · Xh ago' badge reads. Engine-health surface only — no model output.
Top-level shape
{ schema_version, started_at, finished_at, ok, pipelines[], duration_sec }Sample (truncated)
{
"schema_version": 1,
"ok": true,
"duration_sec": 422,
"pipelines": [
{ "name": "seer_earnings.benchmark", "ok": true, "duration_sec": 4.75 },
{ "name": "seer_fomc.ecb.posterior", "ok": true, "duration_sec": 17.3 }
]
}Cross-engine calibration receipts. Walk-forward direction accuracy + Brier across all engines, on resolved historical predictions only. The 'has SEER been right' rollup. No forward picks — past performance only.
Top-level shape
{ generated_at, summary, by_engine, recent_resolved[] }Sample (truncated)
{
"summary": { "n_predictions": 1271, "direction_accuracy": 0.633 },
"by_engine": { "earnings": {...}, "fomc": {...}, ... },
"recent_resolved": [...]
}Partner endpoints
/desk auth required · forward picksThese return the actual model output: which markets to trade, which posteriors are live, which Trump claims diverged from price. Behind basic auth on purpose. Email hello@goseer.ai for partner credentials.
Topics where 2+ engines agree right now. The 'multiple signals pointing the same direction' surface. Forward-looking — gated.
Top-level shape
{ generated_at, topics[], summary }Sample (truncated)
{ "topics": [ { "topic": "Iran", "engines": ["fomc","trump","predmarkets"] } ] }FOMC engine snapshot. Powell + Lagarde + Bailey + (BoJ) cross-asset edges, walk-forward stats, and decision posteriors for the next meetings. Posteriors are forward picks — gated.
Top-level shape
{ powell{}, lagarde{}, bailey{}, boj{}, next_calls[] }Sample (truncated)
{ "powell_next": { "argmax": "hold", "probs": { "hold": 0.84, ... } } }Predmarkets EV scanner snapshot. Top covered markets and the actionable-pick list (≥1pp edge). Backtest summary stays in receipts.json; the live picks are gated.
Top-level shape
{ generated_at, headline, top_events[], mentions_scored[], backtest{} }Sample (truncated)
{ "actionable": [ { "ticker": "...", "side": "BUY NO", "edge_pp": 4.7 } ] }Trump rolling-feed snapshot. Statements + grounded claims + verdict counts + reference-adjusted market moves. Currently v0 (synthesized while live ingest matures). Gated.
Top-level shape
{ schema_version, generation_method, statement_count, claim_count, verdict_counts, statements[] }Sample (truncated)
{ "generation_method": "manual_v0_pre_api_keys", "statement_count": 26 }