Skip to content

Supercharge your AI agents with data from the web and beyond. Building the library for superintelligence. 🔥

docs(php-sdk): frame crawl timeout guidance around runtime behavior

Kris Oeiddd75789ec51
diff --git a/apps/php-sdk/README.md b/apps/php-sdk/README.md
index 31c80cef3..0b346f541 100644
--- a/apps/php-sdk/README.md
+++ b/apps/php-sdk/README.md
@@ -463,9 +463,9 @@ as readable error strings rather than thrown, so agent runs degrade gracefully.
`firecrawl_crawl` waits up to 55 seconds for the crawl to finish. If your
agent runs inside a queued job, keep the crawl limit small or raise the
-worker's job timeout, since a killed job orphans a crawl that is still
-running (and billed) on the server. Extend `FirecrawlCrawl` and override
-`$timeoutSeconds` to change how long it waits.
+worker's job timeout; on timeout the model receives a readable message and
+the crawl may still complete on the server. Extend `FirecrawlCrawl` and
+override `$timeoutSeconds` to change how long it waits.
## License
diff --git a/apps/php-sdk/src/Laravel/Tools/FirecrawlCrawl.php b/apps/php-sdk/src/Laravel/Tools/FirecrawlCrawl.php
index 9bcd8bab0..c5e9eaebe 100644
--- a/apps/php-sdk/src/Laravel/Tools/FirecrawlCrawl.php
+++ b/apps/php-sdk/src/Laravel/Tools/FirecrawlCrawl.php
@@ -16,7 +16,8 @@ class FirecrawlCrawl extends FirecrawlTool
/**
* Extenders can override this to wait longer, but the default keeps the
* crawl inside typical queue worker timeouts (e.g. Laravel's default of
- * 60 seconds) so a billed crawl is not orphaned when the job is killed.
+ * 60 seconds) so the tool returns a readable timeout message instead of
+ * dying with the job.
*/
protected int $timeoutSeconds = 55;