decouple alerts from question/answer

Author: yrobla

src/codegate/api/v1.py

@@ -420,7 +420,9 @@ async def get_workspace_alerts(workspace_name: str) -> List[Optional[v1_models.A
         raise HTTPException(status_code=500, detail="Internal server error")
 
     try:
-        alerts = await dbreader.get_alerts_by_workspace(ws.id, AlertSeverity.CRITICAL.value)
+        alerts = await dbreader.get_alerts_by_workspace_or_prompt_id(
+            workspace_id=ws.id, trigger_category=AlertSeverity.CRITICAL.value
+        )
         prompts_outputs = await dbreader.get_prompts_with_output(ws.id)
         return await v1_processing.parse_get_alert_conversation(alerts, prompts_outputs)
     except Exception:
@@ -576,10 +578,19 @@ async def get_messages_by_prompt_id(
     prompts_outputs = await dbreader.get_prompts_with_output(
         workspace_id=ws.id, prompt_id=prompt_id
     )
+
+    # get all alerts for the prompt
+    alerts = await dbreader.get_alerts_by_workspace_or_prompt_id(
+        workspace_id=ws.id, prompt_id=prompt_id, trigger_category=AlertSeverity.CRITICAL.value
+    )
+    deduped_alerts = await v1_processing.remove_duplicate_alerts(alerts)
     conversations, _ = await v1_processing.parse_messages_in_conversations(prompts_outputs)
     if not conversations:
         raise HTTPException(status_code=404, detail="Conversation not found")
-    return conversations[0]
+
+    conversation = conversations[0]
+    conversation.alerts = deduped_alerts
+    return conversation
 
 
 @v1.get(

src/codegate/api/v1_models.py

@@ -202,7 +202,6 @@ class PartialQuestionAnswer(pydantic.BaseModel):
     partial_questions: PartialQuestions
     answer: Optional[ChatMessage]
     model_token_usage: TokenUsageByModel
-    alerts: List[Alert] = []
 
 
 class Conversation(pydantic.BaseModel):
@@ -216,7 +215,7 @@ class Conversation(pydantic.BaseModel):
     chat_id: str
     conversation_timestamp: datetime.datetime
     token_usage_agg: Optional[TokenUsageAggregate]
-    alerts: List[Alert] = []
+    alerts: Optional[List[Alert]] = []
 
 
 class ConversationSummary(pydantic.BaseModel):

src/codegate/api/v1_processing.py

@@ -202,15 +202,10 @@ async def _get_partial_question_answer(
         model=model, token_usage=token_usage, provider_type=provider
     )
 
-    alerts: List[v1_models.Alert] = [
-        v1_models.Alert.from_db_model(db_alert) for db_alert in row.alerts
-    ]
-
     return PartialQuestionAnswer(
         partial_questions=request_message,
         answer=output_message,
         model_token_usage=model_token_usage,
-        alerts=alerts,
     )
 
 
@@ -374,7 +369,7 @@ async def match_conversations(
     for group in grouped_partial_questions:
         questions_answers: List[QuestionAnswer] = []
         token_usage_agg = TokenUsageAggregate(tokens_by_model={}, token_usage=TokenUsage())
-        alerts: List[v1_models.Alert] = []
+
         first_partial_qa = None
         for partial_question in sorted(group, key=lambda x: x.timestamp):
             # Partial questions don't contain the answer, so we need to find the corresponding
@@ -398,8 +393,6 @@ async def match_conversations(
                 qa = _get_question_answer_from_partial(selected_partial_qa)
                 qa.question.message = parse_question_answer(qa.question.message)
                 questions_answers.append(qa)
-                deduped_alerts = await remove_duplicate_alerts(selected_partial_qa.alerts)
-                alerts.extend(deduped_alerts)
                 token_usage_agg.add_model_token_usage(selected_partial_qa.model_token_usage)
 
         # if we have a conversation with at least one question and answer
@@ -413,7 +406,6 @@ async def match_conversations(
                 chat_id=first_partial_qa.partial_questions.message_id,
                 conversation_timestamp=first_partial_qa.partial_questions.timestamp,
                 token_usage_agg=token_usage_agg,
-                alerts=alerts,
             )
             for qa in questions_answers:
                 map_q_id_to_conversation[qa.question.message_id] = conversation

src/codegate/db/connection.py

@@ -854,8 +854,11 @@ async def get_total_messages_count_by_workspace_id(
                 logger.error(f"Failed to fetch message count. Error: {e}")
                 return 0  # Return 0 in case of failure
 
-    async def get_alerts_by_workspace(
-        self, workspace_id: str, trigger_category: Optional[str] = None
+    async def get_alerts_by_workspace_or_prompt_id(
+        self,
+        workspace_id: str,
+        prompt_id: Optional[str] = None,
+        trigger_category: Optional[str] = None,
     ) -> List[Alert]:
         sql = text(
             """
@@ -874,6 +877,10 @@ async def get_alerts_by_workspace(
         )
         conditions = {"workspace_id": workspace_id}
 
+        if prompt_id:
+            sql = text(sql.text + " AND a.prompt_id = :prompt_id")
+            conditions["prompt_id"] = prompt_id
+
         if trigger_category:
             sql = text(sql.text + " AND a.trigger_category = :trigger_category")
             conditions["trigger_category"] = trigger_category

src/codegate/db/models.py

@@ -352,4 +352,3 @@ class GetMessagesRow(BaseModel):
     output_tokens: Optional[int]
     input_cost: Optional[float]
     output_cost: Optional[float]
-    alerts: List[Alert] = []