updates

Author: arvindrk

fern/observability/evals-quickstart.mdx

@@ -576,7 +576,7 @@ For complex validation criteria beyond pattern matching, use AI-powered judges t
 ```
 You are an LLM-Judge. Evaluate ONLY the last assistant message in the mock conversation: {{messages[-1]}}.
 
-Include the full conversation history for context: {{messages[0:-1]}}
+Include the full conversation history for context: {{messages}}
 
 Decision rule:
 - PASS if ALL "pass criteria" are satisfied AND NONE of the "fail criteria" are triggered.
@@ -596,6 +596,12 @@ Output format: respond with exactly one word: pass or fail
 - No additional text
 ```
 
+<Note>
+**Template variables:**
+- `{{messages}}` - The entire conversation history (all messages exchanged)
+- `{{messages[-1]}}` - The last assistant message only
+</Note>
+
 ### Example: Evaluate helpfulness and tone
 
 <Tabs>
@@ -630,7 +636,7 @@ curl -X POST "https://api.vapi.ai/eval" \
             "model": "gpt-4o",
             "messages": [{
               "role": "system",
-              "content": "You are an LLM-Judge. Evaluate ONLY the last assistant message: {{messages[-1]}}.\n\nInclude context: {{messages[0:-1]}}\n\nDecision rule:\n- PASS if ALL pass criteria are met AND NO fail criteria are triggered.\n- Otherwise FAIL.\n\nPass criteria:\n- Response acknowledges the user request\n- Response offers specific help or next steps\n- Tone is professional and friendly\n\nFail criteria (any triggers FAIL):\n- Response is rude or dismissive\n- Response ignores the user request\n- Response provides no actionable information\n\nOutput format: respond with exactly one word: pass or fail"
+              "content": "You are an LLM-Judge. Evaluate ONLY the last assistant message: {{messages[-1]}}.\n\nInclude context: {{messages}}\n\nDecision rule:\n- PASS if ALL pass criteria are met AND NO fail criteria are triggered.\n- Otherwise FAIL.\n\nPass criteria:\n- Response acknowledges the user request\n- Response offers specific help or next steps\n- Tone is professional and friendly\n\nFail criteria (any triggers FAIL):\n- Response is rude or dismissive\n- Response ignores the user request\n- Response provides no actionable information\n\nOutput format: respond with exactly one word: pass or fail"
             }]
           }
         }
@@ -1366,11 +1372,13 @@ Run multiple evals sequentially to validate all greeting scenarios.
   </Card>
 
 {" "}
+
 <Card title="Assistants guide" icon="robot" href="/assistants/quickstart">
   Create and configure assistants to test
 </Card>
 
 {" "}
+
 <Card title="Tools documentation" icon="wrench" href="/tools/custom-tools">
   Build custom tools and validate their behavior
 </Card>