Bump HTTP procedure timeouts (500ms/10s → 30s/180s) and fix docs

crates/core/src/host/instance_env.rs

@@ -970,12 +970,16 @@ impl InstanceEnv {
 
 /// Default timeout for HTTP requests performed by [`InstanceEnv::http_request`].
 ///
-/// Value chosen arbitrarily by pgoldman 2025-11-18, based on little more than a vague guess.
-const HTTP_DEFAULT_TIMEOUT: Duration = Duration::from_millis(500);
+/// Applied when the module does not specify a timeout.
+/// 30 seconds is generous enough for most external API calls (including LLM APIs)
+/// without silently hanging forever on a broken endpoint.
+const HTTP_DEFAULT_TIMEOUT: Duration = Duration::from_secs(30);
 /// Maximum timeout for HTTP requests performed by [`InstanceEnv::http_request`].
 ///
 /// If the user requests a timeout longer than this, we will clamp to this value.
-const HTTP_MAX_TIMEOUT: Duration = Duration::from_secs(10);
+/// 180 seconds accommodates long-running LLM and AI API calls,
+/// which routinely take 30-120 seconds for complex requests.
+const HTTP_MAX_TIMEOUT: Duration = Duration::from_secs(180);
 const BLOCKED_HTTP_ADDRESS_ERROR: &str = "refusing to connect to private or special-purpose addresses";
 
 struct FilteredDnsResolver;

docs/docs/00200-core-concepts/00200-functions/00400-procedures.md

@@ -873,15 +873,15 @@ SPACETIMEDB_PROCEDURE(Unit, get_request_with_short_timeout, ProcedureContext ctx
 }
 ```
 
-:::note
-All timeouts are clamped to a maximum of 500ms by the host.
-:::
-
 Procedures can't send requests at the same time as holding open a [transaction](#accessing-the-database).
 
 </TabItem>
 </Tabs>
 
+:::note
+If no timeout is specified, HTTP requests default to 30 seconds. User-specified timeouts are clamped to a maximum of 180 seconds by the host.
+:::
+
 ## Calling Reducers from Procedures
 
 Procedures can call reducers by invoking them within a transaction block. The reducer function runs within the transaction context: