⚠️ Potential issue

Incorrect import for WorkflowFailureError

WorkflowFailureError is exposed from temporalio.exceptions, not temporalio.client. This will raise ImportError at runtime.

-from temporalio.client import WorkflowFailureError
+from temporalio.exceptions import WorkflowFailureError

In tasks/hivemind/query_data_sources.py around line 12, the code imports
WorkflowFailureError from temporalio.client which is incorrect; change the
import to import WorkflowFailureError from temporalio.exceptions (i.e., replace
the temporalio.client import with temporalio.exceptions) so the correct symbol
is imported and ImportError at runtime is avoided.

🛠️ Refactor suggestion

Ensure workflow ID uniqueness when workflow_id is absent

Using id=f"hivemind-query-{community_id}-{workflow_id}" will collide across requests when workflow_id is None, leading to WorkflowAlreadyStarted-like errors. Fallback to a random suffix when no workflow_id is provided.

-                id=f"hivemind-query-{self.community_id}-{self.workflow_id}",
+                id=f"hivemind-query-{self.community_id}-{self.workflow_id or os.urandom(8).hex()}",

Committable suggestion skipped: line range outside the PR's diff.

In tasks/hivemind/query_data_sources.py around lines 47 to 55, the workflow id
string uses self.workflow_id directly which can be None and cause collisions;
update the code to compute a unique workflow_id_part (use self.workflow_id if
truthy otherwise generate a short random suffix such as a UUID4 or
timestamp+random) and then use
id=f"hivemind-query-{self.community_id}-{workflow_id_part}" when calling
client.execute_workflow so each started workflow without an explicit workflow_id
gets a unique id to avoid WorkflowAlreadyStarted errors.

🛠️ Refactor suggestion

Don’t log exc_info for non-exceptions

These branches handle failure-shaped payloads, not exceptions. exc_info=True prints a spurious “NoneType: None” stack trace. Remove exc_info here.

-            logging.error(f"WorkflowFailureError: {result} for workflow {self.workflow_id}", exc_info=True)
+            logging.error(f"WorkflowFailureError-shaped payload for workflow {self.workflow_id}: {result}")
...
-            logging.error(f"WorkflowFailureError: {result} for workflow {self.workflow_id}", exc_info=True)
+            logging.error(f"WorkflowFailureError-shaped payload for workflow {self.workflow_id}: {result}")

In tasks/hivemind/query_data_sources.py around lines 62 to 70, the logging.error
calls for failure-shaped payloads include exc_info=True which prints a spurious
stack trace for non-exceptions; remove the exc_info=True argument from both
logging.error calls so the message is logged without exception/stack trace
information.

💡 Verification agent

if not isinstance(result, str):
    result = json.dumps(result, ensure_ascii=False)
return result

#!/bin/bash
# Locate the function definition and show its signature
rg -n "def " tasks/hivemind/query_data_sources.py

# Show the first 200 lines to capture imports and the method signature
sed -n '1,200p' tasks/hivemind/query_data_sources.py

Ensure query always returns a string or update its signature

The async def query(self, …) -> str | None currently returns whatever the workflow yields—often a dict—which violates the declared return type and will break the get_rag_answer tool (it expects a str). You should either serialize non‐string results or widen the type hint.

Locations to update in tasks/hivemind/query_data_sources.py:

• After the failure checks (around line 70), coerce dicts to JSON:

       # Normalize Temporal failure-shaped responses …
       if isinstance(result, dict) and ("workflowExecutionFailedEventAttributes" in result or "failure" in result):
           logging.error(…)
           return None
       if isinstance(result, str) and "workflowExecutionFailedEventAttributes" in result:
           logging.error(…)
           return None
  

•  # Ensure successful payloads are returned as strings
  

•  if not isinstance(result, str):
  

•      import json
  

•      result = json.dumps(result, ensure_ascii=False)
  
   return result
  

• Or, if you intend to surface structured data, change the signature to:

  async def query(self, query: str) -> dict | str | None:

  and update the get_rag_answer wrapper accordingly.

In tasks/hivemind/query_data_sources.py around lines 62–71, the function
annotated as async def query(...) -> str | None sometimes returns dicts (or
other non-string types), violating the signature and breaking callers; after the
existing failure checks, coerce non-string results to strings (e.g., import json
and if isinstance(result, dict) set result = json.dumps(result), and for other
non-str types convert with str(result)) so the function always returns a str or
None, or alternatively change the signature to async def query(self, query: str)
-> dict | str | None and update get_rag_answer to accept structured results—pick
one approach and apply the corresponding code and type-hint changes.