fix: add configurable HTTP timeout and preserve error context

[skyc1e]

Summary

Fixes #273.

The default httpx timeout (~5s) causes silent order duplication under load: requests time out client-side, the SDK raises a generic PolyApiException, callers assume failure and retry, but the original order was already matched server-side. Multiple users reported duplicate fills from this behavior.

Changes

• Added optional timeout parameter to ClobClient.__init__() — when provided, reconfigures the module-level httpx client with the specified timeout in seconds
• Preserved the original httpx.RequestError via exception chaining (raise ... from exc) so callers can distinguish timeouts from DNS failures, connection resets, etc.
• Fully backward compatible: default behavior unchanged when timeout is not specified

Usage

client = ClobClient(
    host="https://clob.polymarket.com",
    key=key,
    chain_id=137,
    timeout=30,
)

Test plan

• Verified default behavior unchanged (no timeout arg = httpx default)
• Verified timeout=30 reconfigures the client correctly
• Verified exception chaining preserves original error type and message

---

Note

Medium Risk
Changes HTTP client configuration and error propagation for all SDK requests, which could alter runtime behavior (timeouts) and exception handling in consumers.

Overview
Adds an optional timeout argument to ClobClient.__init__ that reconfigures the module-level httpx client via new set_http_timeout, allowing callers to override default request timeouts.

Improves HTTP error reporting by including the original httpx.RequestError message and chaining the exception (raise ... from exc) so consumers can distinguish timeouts and connection failures.

^{Written by Cursor Bugbot for commit b69208f. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Old HTTP client not closed, leaking connections

Medium Severity

set_http_timeout replaces the module-level _http_client with a new httpx.Client without calling .close() on the old one. Since the client is created with http2=True, the old instance holds open HTTP/2 connections and associated resources (sockets, TLS state) that are never properly released, causing a resource leak.

 

[cursor]

Per-instance timeout parameter actually mutates global state

High Severity

The timeout parameter on ClobClient.__init__ suggests per-instance configuration, but set_http_timeout replaces a module-level global _http_client shared by all instances. Creating a second ClobClient with a different (or no) timeout silently overrides the first client's timeout. Ironically, this can reintroduce the exact duplicate-order problem the PR aims to fix — a user sets timeout=30 on one client, then another component instantiates a default client, resetting the timeout to httpx's default.

Additional Locations (1)

• py_clob_client/http_helpers/helpers.py#L21-L27