Move eval 03 out of easy (#912)

Author: aantn

pyproject.toml

@@ -101,7 +101,6 @@ markers = [
     "synthetic: Tests using synthetic data",
     "network: Tests requiring network connectivity",
     "runbooks: Tests involving runbook functionality",
-    "kafka: Tests involving Kafka functionality",
     "misleading-history: Tests with misleading historical data",
     "chain-of-causation: Tests involving chain-of-causation analysis",
     "slackbot: Tests involving Slack bot functionality",
@@ -119,8 +118,9 @@ markers = [
     "toolset-limitation: Tests that cannot be solved no matter how smart the model, unless we improve the underlying toolsets themselves",
     "ask-for-clarification: Tests where Holmes should ask the user for clarification",
     "database: Tests involving database interactions",
-    "datadog: DataDog toolset",
-    "traces: Tests where the ai is expected to find the solution using the traces"
+    "traces: Tests where the ai is expected to find the solution using the traces",
+    "answer-given-wrong-place: LLM gave the answer in intermediate steps, not final answer, so its harder for the user to see the answer",
+    "datadog: DataDog toolset"
 ]
 
 addopts = [

tests/llm/fixtures/test_ask_holmes/03_what_is_the_command_to_port_forward/test_case.yaml

@@ -6,7 +6,8 @@ expected_output:
   - Should NOT just say "find grafana in your cluster" or give generic instructions
 tags:
   - kubernetes
-  - easy
+  - medium
+  - answer-given-wrong-place # LLM gave the answer in intermediate steps, not final answer, so harder for user to see
 before_test: |
   kubectl create namespace app-03
   cat <<EOF | kubectl apply -f -