Merge PR #616: fix: retry with rebuilt payload after compression

Authored by tripledoublev.

After context compression on 413/400 errors, the inner retry loop was
reusing the stale pre-compression api_messages payload. Fix breaks out
of the inner retry loop so the outer loop rebuilds api_messages from
the now-compressed messages list. Adds regression test verifying the
second request actually contains the compressed payload.

run_agent.py

@@ -3374,6 +3374,7 @@ class AIAgent:
             max_compression_attempts = 3
             codex_auth_retry_attempted = False
             nous_auth_retry_attempted = False
+            restart_with_compressed_messages = False
 
             finish_reason = "stop"
             response = None  # Guard against UnboundLocalError if all retries fail
@@ -3707,7 +3708,8 @@ class AIAgent:
                         if len(messages) < original_len:
                             print(f"{self.log_prefix}   🗜️  Compressed {original_len} → {len(messages)} messages, retrying...")
                             time.sleep(2)  # Brief pause between compression retries
-                            continue  # Retry with compressed messages
+                            restart_with_compressed_messages = True
+                            break
                         else:
                             print(f"{self.log_prefix}❌ Payload too large and cannot compress further.")
                             logging.error(f"{self.log_prefix}413 payload too large. Cannot compress further.")
@@ -3775,7 +3777,8 @@ class AIAgent:
                             if len(messages) < original_len:
                                 print(f"{self.log_prefix}   🗜️  Compressed {original_len} → {len(messages)} messages, retrying...")
                             time.sleep(2)  # Brief pause between compression retries
-                            continue  # Retry with compressed messages or new tier
+                            restart_with_compressed_messages = True
+                            break
                         else:
                             # Can't compress further and already at minimum tier
                             print(f"{self.log_prefix}❌ Context length exceeded and cannot compress further.")
@@ -3862,6 +3865,11 @@ class AIAgent:
             if interrupted:
                 break
 
+            if restart_with_compressed_messages:
+                api_call_count -= 1
+                self.iteration_budget.refund()
+                continue
+
             # Guard: if all retries exhausted without a successful response
             # (e.g. repeated context-length errors that exhausted retry_count),
             # the `response` variable is still None. Break out cleanly.

tests/test_413_compression.py

@@ -234,6 +234,55 @@ class TestHTTP413Compression:
         mock_compress.assert_called_once()
         assert result["completed"] is True
 
+    def test_context_length_retry_rebuilds_request_after_compression(self, agent):
+        """Retry must send the compressed transcript, not the stale oversized payload."""
+        err_400 = Exception(
+            "Error code: 400 - {'error': {'message': "
+            "\"This endpoint's maximum context length is 128000 tokens. "
+            "Please reduce the length of the messages.\"}}"
+        )
+        err_400.status_code = 400
+        ok_resp = _mock_response(content="Recovered after real compression", finish_reason="stop")
+
+        request_payloads = []
+
+        def _side_effect(**kwargs):
+            request_payloads.append(kwargs)
+            if len(request_payloads) == 1:
+                raise err_400
+            return ok_resp
+
+        agent.client.chat.completions.create.side_effect = _side_effect
+
+        prefill = [
+            {"role": "user", "content": "previous question"},
+            {"role": "assistant", "content": "previous answer"},
+        ]
+
+        with (
+            patch.object(agent, "_compress_context") as mock_compress,
+            patch.object(agent, "_persist_session"),
+            patch.object(agent, "_save_trajectory"),
+            patch.object(agent, "_cleanup_task_resources"),
+        ):
+            mock_compress.return_value = (
+                [{"role": "user", "content": "compressed summary"}],
+                "compressed prompt",
+            )
+            result = agent.run_conversation("hello", conversation_history=prefill)
+
+        assert result["completed"] is True
+        assert len(request_payloads) == 2
+        assert len(request_payloads[1]["messages"]) < len(request_payloads[0]["messages"])
+        assert request_payloads[1]["messages"][0] == {
+            "role": "system",
+            "content": "compressed prompt",
+        }
+        assert request_payloads[1]["messages"][1] == {
+            "role": "user",
+            "content": "compressed summary",
+        }
+
     def test_413_cannot_compress_further(self, agent):
         """When compression can't reduce messages, return partial result."""
         err_413 = _make_413_error()