fix(autofix): Fix rethink bug in plan+code due to pre-expanded files (#…

…1373)

The memory truncation logic upon a rethink always kept only 1 memory
item if you did it from before the first insight card. This broke
certain cases in the plan+code step since it can have multiple initial
messages. #1371

This PR fixes that by tracking how many items make up the "initial
prompt".

(also tacking on a small prompt tweak for insight sharing that seems to
help with #1372

src/seer/automation/autofix/components/coding/component.py

@@ -111,7 +111,11 @@ def invoke(self, request: CodingRequest) -> CodingOutput | None:
             )
 
             has_tools = True
-            if isinstance(request.root_cause_and_fix, RootCauseAnalysisItem):
+
+            if (
+                isinstance(request.root_cause_and_fix, RootCauseAnalysisItem)
+                and not request.initial_memory
+            ):
                 expanded_files_messages = []
                 relevant_files = [
                     {"file_path": context.snippet.file_path, "repo_name": context.snippet.repo_name}
@@ -144,6 +148,9 @@ def invoke(self, request: CodingRequest) -> CodingOutput | None:
                         expanded_files_messages.append(agent_message)
                         expanded_files_messages.append(user_message)
 
+                with self.context.state.update() as cur:
+                    cur.steps[-1].initial_memory_length = len(expanded_files_messages) + 1
+
                 if expanded_files_messages and not request.initial_memory:
                     is_obvious = IsFixObviousComponent(self.context).invoke(
                         IsFixObviousRequest(

src/seer/automation/autofix/components/insight_sharing/component.py

@@ -30,6 +30,7 @@ def format_step_one(
 
             Write the next under-20-word conclusion in the chain of thought based on the notes below, or if there is no good conclusion to add, return <NO_INSIGHT/>. The criteria for a good conclusion are that it should be a large, novel jump in insights, not similar to any item in the existing chain of thought, it should be a complete conclusion after some meaty analysis, not a plan of what to analyze next, and it should be valuable for {task_description}. It should also be very concrete, to-the-point, and specific. Every item in the chain of thought should read like a chain that clearly builds off of the previous step. If you can't find a conclusion that meets ALL of these criteria, return <NO_INSIGHT/>.
 
+            NOTES TO ANALYZE:
             {latest_thought}"""
         ).format(
             task_description=task_description,

src/seer/automation/autofix/models.py

@@ -169,6 +169,7 @@ def is_valid_uuid(uuid_string: str) -> bool:
 class DefaultStep(BaseStep):
     type: Literal[StepType.DEFAULT] = StepType.DEFAULT
     insights: list[InsightSharingOutput] = []
+    initial_memory_length: int = 1
 
 
 class RootCauseStep(BaseStep):

src/seer/automation/autofix/tasks.py

@@ -338,7 +338,7 @@ def truncate_memory_to_match_insights(memory: list[Message], step: DefaultStep):
             truncated_memory = new_memory
             break
     if not step.insights:
-        truncated_memory = memory[:1]
+        truncated_memory = memory[: step.initial_memory_length]
     return truncated_memory if truncated_memory else memory
 
 

tests/automation/autofix/test_autofix_tasks.py

@@ -410,6 +410,7 @@ def test_restart_from_point_with_feedback_root_cause_step(
         mock_step = MagicMock(spec=DefaultStep)
         mock_step.key = "root_cause_analysis_processing"
         mock_step.insights = []
+        mock_step.initial_memory_length = 1
         mock_state.get.return_value.steps = [mock_step]
         mock_state.get.return_value.run_id = 123
         mock_state.update.return_value.__enter__.return_value.steps = [mock_step]
@@ -437,6 +438,7 @@ def test_truncate_memory_to_match_insights(self):
             MagicMock(spec=InsightSharingOutput, generated_at_memory_index=4),
             MagicMock(spec=InsightSharingOutput, generated_at_memory_index=-1),
         ]
+        mock_step.initial_memory_length = 1
 
         # Create a mock memory list
         memory = [