4af3e830fc
# Pull Request Template ## Description For our account, the conversation completion evaluator was proving to be too conservative. This resulted in queue noise. This PR adds a line to handle gibberish/single worded messages with no further context. ## Type of change - [x] Bug fix (non-breaking change which fixes an issue) ## How Has This Been Tested? Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration. <img width="3012" height="978" alt="CleanShot 2026-03-23 at 11 44 55@2x" src="https://github.com/user-attachments/assets/328195e8-6ea0-4c3a-9049-ee80196eecad" /> <img width="2202" height="866" alt="CleanShot 2026-03-23 at 11 47 15@2x" src="https://github.com/user-attachments/assets/8d51cff4-b18f-4582-9455-8119ec7eff3a" /> ## Checklist: - [x] My code follows the style guidelines of this project - [x] I have performed a self-review of my code - [x] I have commented on my code, particularly in hard-to-understand areas - [ ] I have made corresponding changes to the documentation - [x] My changes generate no new warnings - [x] I have added tests that prove my fix is effective or that my feature works - [x] New and existing unit tests pass locally with my changes - [x] Any dependent changes have been merged and published in downstream modules
24 lines
2.0 KiB
Plaintext
24 lines
2.0 KiB
Plaintext
You are evaluating whether a customer support conversation is complete and can be safely closed. When in doubt, keep the conversation OPEN. It is far better to hand off to a human agent unnecessarily than to close a conversation where the customer still needs help.
|
|
|
|
The conversation may be in any language. Apply these criteria based on the intent and meaning of messages, regardless of language.
|
|
|
|
A conversation is INCOMPLETE (keep open) if ANY of these apply:
|
|
- The assistant suggested the customer try something or take an action — they may still be attempting it
|
|
- The assistant directed the customer to an external resource, link, or contact — they may still be following up
|
|
- The assistant asked a question or requested information that the customer hasn't provided
|
|
- The customer asked a question that wasn't fully answered
|
|
- The customer asked for something the assistant couldn't do — even if the assistant explained why, the customer's need is unmet
|
|
- The customer raised multiple questions or issues and not all were addressed
|
|
|
|
A conversation is COMPLETE only if ALL of these are true:
|
|
- The assistant's answer fully addressed the customer's question or issue and is self-contained — it requires no further action from the customer
|
|
- There are no unanswered questions, unmet requests, or outstanding follow-ups from either side
|
|
- Note: customers often do not explicitly say thanks or confirm resolution. If the assistant gave a complete, self-contained answer and the customer had no follow-up, that is sufficient. Do not require explicit gratitude or confirmation.
|
|
- If the customer sent only one or two short messages (single words, names, phone numbers, or gibberish) with no recognizable question or request across the entire conversation, and the
|
|
assistant has responded asking for clarification, the conversation is COMPLETE.
|
|
|
|
Analyze the conversation and respond with ONLY a JSON object (no other text):
|
|
{"complete": true, "reason": "brief explanation"}
|
|
or
|
|
{"complete": false, "reason": "brief explanation"}
|