From 5b57bf3dd048f66f3b83e83d6044d9255b9fea7b Mon Sep 17 00:00:00 2001
From: hermes <hermes@timmy.local>
Date: Sun, 15 Mar 2026 12:08:30 -0400
Subject: [PATCH] [loop-cycle-50] fix: agent retry uses exponential backoff
 instead of fixed 1s delay (#174) (#181)

---
 src/timmy/agents/base.py        |  2 +-
 tests/timmy/test_agent_retry.py | 23 ++++++++++++++++++-----
 2 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/src/timmy/agents/base.py b/src/timmy/agents/base.py
index a4d64dac..0ea9ea87 100644
--- a/src/timmy/agents/base.py
+++ b/src/timmy/agents/base.py
@@ -145,7 +145,7 @@ class BaseAgent(ABC):
                         max_retries,
                         exc,
                     )
-                    await asyncio.sleep(1)
+                    await asyncio.sleep(min(2 ** (attempt - 1), 8))
                 else:
                     logger.error(
                         "Agent run failed after %d attempts: %s",
diff --git a/tests/timmy/test_agent_retry.py b/tests/timmy/test_agent_retry.py
index c2511f3c..dfa5d1e1 100644
--- a/tests/timmy/test_agent_retry.py
+++ b/tests/timmy/test_agent_retry.py
@@ -1,6 +1,6 @@
 """Tests for agent retry logic on transient errors."""
 
-from unittest.mock import MagicMock
+from unittest.mock import MagicMock, patch
 
 import httpx
 import pytest
@@ -42,11 +42,16 @@ async def test_run_retries_on_transient_error(sub_agent):
     sub_agent.agent.run = mock_run
 
     # Act
-    result = await sub_agent.run("test message")
+    with patch("timmy.agents.base.asyncio.sleep") as mock_sleep:
+        result = await sub_agent.run("test message")
 
     # Assert
     assert result == "Success after retries"
     assert call_count == 3  # 2 failures + 1 success
+    # Verify exponential backoff: attempt 1 = 1s, attempt 2 = 2s
+    assert mock_sleep.call_count == 2
+    mock_sleep.assert_any_call(1)
+    mock_sleep.assert_any_call(2)
 
 
 @pytest.mark.asyncio
@@ -56,11 +61,16 @@ async def test_run_exhausts_retries(sub_agent):
     sub_agent.agent.run.side_effect = Exception("Ollama 500 error: XML parse error")
 
     # Act & Assert
-    with pytest.raises(Exception, match="Ollama 500 error: XML parse error"):
-        await sub_agent.run("test message")
+    with patch("timmy.agents.base.asyncio.sleep") as mock_sleep:
+        with pytest.raises(Exception, match="Ollama 500 error: XML parse error"):
+            await sub_agent.run("test message")
 
     # Should have been called 3 times (max retries)
     assert sub_agent.agent.run.call_count == 3
+    # Verify exponential backoff: attempt 1 = 1s, attempt 2 = 2s
+    assert mock_sleep.call_count == 2
+    mock_sleep.assert_any_call(1)
+    mock_sleep.assert_any_call(2)
 
 
 @pytest.mark.asyncio
@@ -114,10 +124,13 @@ async def test_run_logs_retry_attempts(sub_agent, caplog):
         sub_agent.agent.run = mock_run
 
         # Act
-        result = await sub_agent.run("test message")
+        with patch("timmy.agents.base.asyncio.sleep") as mock_sleep:
+            result = await sub_agent.run("test message")
 
         # Assert
         assert result == "Success"
         assert call_count == 2  # 1 failure + 1 success
         assert "Agent run failed on attempt 1/3" in caplog.text
         assert "Retrying..." in caplog.text
+        # Verify sleep was called with 1s for first attempt
+        mock_sleep.assert_called_once_with(1)