You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
allow_unknown_host is partially covered by the prose under URL validation (lines 66-77) but is not in the at-a-glance reference table users actually skim. The other three predate v2.5 but have never appeared in the reference table.
Users reaching for total_timeout (the wall-clock retry budget #193 introduced) or ws_ping_interval (heartbeat tuning) currently have to discover them by reading the dataclass source — defeats the purpose of the doc page.
Suggested fix
Append these rows to the table at docs/configuration.md:46:
|`total_timeout`|`None`| Wall-clock cap (seconds) across the whole request including all retries. `None` = unbounded; cap with the per-attempt `timeout` only. See [Retries & idempotency](retries.md). ||`ws_ping_interval`|`20.0`| Seconds between WebSocket pings (forwarded to the `websockets` library). ||`ws_close_timeout`|`5.0`| Seconds to wait for a clean WebSocket close handshake before forcing the socket shut. ||`allow_unknown_host`|`False`| When `False`, the SDK rejects `base_url` / `ws_base_url` whose host is not in the known-Kalshi allowlist (`api.elections.kalshi.com`, `demo-api.kalshi.co`) or a loopback host. Opt out via the field or `KALSHI_ALLOW_UNKNOWN_HOST=1` if you target a custom proxy. (Added in v2.5.0 #250.) |
Optionally tighten the URL validation section to point at the new allow_unknown_host row instead of repeating its description.
Activity
release: 2.6.0 — post-v2.5 independent reviewer SDK audit closure (#295-