Skip to main content
← Back to the current changelog

​
2026-05-01 — Cursor pagination on /v2/onchain/*/trades

The two trade-history endpoints now support cursor pagination in addition to offset:
  • GET /v2/onchain/markets/{token_id}/trades?cursor=<lastTs>:<lastId>
  • GET /v2/onchain/wallets/{address}/trades?cursor=<lastTs>:<lastId>
Why it matters: offset pagination slows down the deeper you page — page 500 means the database walks 50,000 rows of wasted work to find your starting point. Cursor pagination loads page 1000 just as fast as page 1 — typically ~1 second per page, regardless of how deep you go. How to use:
  1. First page: pass ?cursor= (empty value) instead of ?offset=0.
  2. Response includes a pagination block with cursor and has_more.
  3. Subsequent pages: pass that cursor back as ?cursor=<value>.
  4. Stop when has_more is false.
Available on every paid tier — no Growth restriction. Compatibility: purely additive. The existing ?offset= mode is unchanged and remains fully supported. Existing integrations don’t need to change anything — switch to cursor when you next touch the integration. See the updated docs for full examples: Trade History (Market) · Trade History (Wallet).

​
2026-05-01 — Bulk trade-export endpoints (Growth plan and above)

Two new endpoints return the entire trade history for a market token or wallet in a single response — no client-side pagination required:
  • GET /v2/onchain/markets/{token_id}/trades/all
  • GET /v2/onchain/wallets/{address}/trades/all
The endpoint walks the full history for you. Optional from and to query parameters narrow the window. Limits:
  • Hard cap 250,000 trades per call. Beyond that, response includes "partial": true with "partial_reason": "hard_cap_250000".
  • Wall-clock budget 180 seconds. First cold-cache calls on busy markets/heavy wallets may take 1-3 minutes. Subsequent calls hit Redis cache and return in under 1 second.
  • Typical payload 1-50 MB. Worst case (capped): 100-200 MB JSON.
  • Cache TTL 5 minutes per (scope, from, to) combination. Partial results are not cached.
Tier requirement: Growth plan ($200/mo) or above. Starter and free tier requests receive 403. The bulk endpoints are gated to paid plans because of their large query and payload sizes. Compatibility: purely additive. The standard paginated /trades endpoints are unchanged and remain available on all paid tiers. Updated the request timeout to 180 seconds so cold-cache calls complete cleanly.

​
2026-05-01 — Backtest Copy PnL: new total_realized_pnl_usdc field

Added total_realized_pnl_usdc to /v2/copy-pnl/{wallet} responses. This is the signed sum of WAVG-realized PnL across every position the matcher touched in the window — the closest single number to “did this wallet make money trading?” and the recommended primary “PnL” field for customer-facing UI. Existing actual_pnl_usdc is unchanged but is now documented as a cashflow metric (USDC delta in/out of the wallet over the window), which is intentionally different from realized PnL — see the endpoint docs for the worked example. Also fixed a long-standing bug in the cashflow accounting where neg-risk-conversion events were over-counting settlement_in (treating share counts as USDC). NRC-active wallets see corrected numbers across all four periods (14d, 30d, 60d, 180d) on the next BYOB cycle refresh.