docs(sdk): document run-scoped LLM headers

sdk/guides/observability.mdx

@@ -140,6 +140,37 @@ Each conversation gets its own session ID (the conversation UUID), allowing you
 
 In `tool.execute`, the tool calls are traced individually, such as `bash`, `file_editor`, or `task_tracker`.
 
+### Correlate LLM Requests With Application Requests
+
+Pass `llm_extra_headers` to attach application metadata to every LLM call made
+during one conversation run:
+
+```python icon="python" wrap
+conversation.send_message("Investigate the failing deployment")
+conversation.run(
+    llm_extra_headers={"X-Request-ID": request_id},
+)
+```
+
+The same option works with `await conversation.arun(...)` and with
+`RemoteConversation`. For direct Agent Server clients, include it in the run request:
+
+```bash icon="terminal" wrap
+curl -X POST "${AGENT_SERVER_URL}/api/conversations/${CONVERSATION_ID}/run" \
+  -H "Content-Type: application/json" \
+  -d '{"llm_extra_headers":{"X-Request-ID":"request-123"}}'
+```
+
+The headers apply to the main agent, retries, condensation, and sub-agent LLM calls
+started by that run. They override static `LLM.extra_headers` with the same name, but
+SDK-managed headers are preserved. Run-scoped headers are not persisted and do not
+affect later runs.
+
+<Note>
+Header values are sent to the configured LLM provider. Treat them as sensitive and
+avoid logging them or placing secrets in correlation headers.
+</Note>
+
 ## Configuration Reference
 
 ### Environment Variables