Customer Agent: Termination Handling
Tracks occurrences of the agent freezing, hanging up abruptly, crashing, or cutting off the conversation early.
result = evaluator.evaluate(
eval_templates="customer_agent_termination_handling",
inputs={
"conversation": "User: I need help with my subscription.\nAgent: Sure, I can help with that. What would you like to know?\nUser: How do I cancel?\nAgent: [call ended]"
},
model_name="turing_flash"
)
print(result.eval_results[0].output)
print(result.eval_results[0].reason)import { Evaluator, Templates } from "@future-agi/ai-evaluation";
const evaluator = new Evaluator();
const result = await evaluator.evaluate(
"customer_agent_termination_handling",
{
conversation: "User: I need help with my subscription.\nAgent: Sure, I can help with that. What would you like to know?\nUser: How do I cancel?\nAgent: [call ended]"
},
{
modelName: "turing_flash",
}
);
console.log(result); | Input | |||
|---|---|---|---|
| Required Input | Type | Description | |
conversation | string | The full conversation history between the customer and agent |
| Output | ||
|---|---|---|
| Field | Description | |
| Result | Returns one of: never, occasionally, frequently, or always — indicating how often termination issues occur | |
| Reason | Provides a detailed explanation of the termination handling assessment |
What to Do When Termination Issues are Detected
- Investigate system logs for errors or timeouts around the flagged moments
- Add graceful error handling to prevent abrupt disconnections
- Implement conversation state persistence to recover from crashes
- Test edge cases that may trigger unexpected termination
Comparing Termination Handling with Similar Evals
- Customer Agent: Interruption Handling: Termination Handling tracks abrupt or unexpected call endings, while Interruption Handling evaluates how the agent recovers when the user interrupts mid-response.
- Conversation Resolution: Termination Handling detects technical failures, while Conversation Resolution evaluates whether the user’s query was fully addressed before the conversation ended.
Was this page helpful?