wehub-resource-sync
48 lines
1.6 KiB
Python
48 lines
1.6 KiB
Python
import textwrap
|
|
from typing import List
|
|
|
|
|
|
class SimulationGraphTemplate:
|
|
@staticmethod
|
|
def classify_edge(assistant_reply: str, choices: List[str]) -> str:
|
|
"""Prompt the simulator model to pick the index of the outgoing edge
|
|
that best describes the assistant's reply. A "None of the above"
|
|
option is appended automatically.
|
|
"""
|
|
numbered_choices = "\n".join(
|
|
f" {i}) {choice}" for i, choice in enumerate(choices, start=1)
|
|
)
|
|
none_index = len(choices) + 1
|
|
prompt = textwrap.dedent(
|
|
f"""You are routing a simulated conversation. The assistant just said:
|
|
<<<
|
|
{assistant_reply}
|
|
>>>
|
|
|
|
Which of the following best describes the assistant's reply?
|
|
{numbered_choices}
|
|
{none_index}) None of the above
|
|
|
|
Pick exactly one option. Respond with the option's number.
|
|
|
|
IMPORTANT: The output must be formatted as a JSON object with two keys:
|
|
- `index`: the 1-based number of the option you chose, OR `null` if you chose "None of the above" (option {none_index}).
|
|
- `reason`: a short rationale (one sentence).
|
|
|
|
Example JSON Output for a match:
|
|
{{
|
|
"index": 1,
|
|
"reason": "The reply explicitly approved the refund."
|
|
}}
|
|
|
|
Example JSON Output for no match:
|
|
{{
|
|
"index": null,
|
|
"reason": "The reply asked a clarifying question that does not fit any option."
|
|
}}
|
|
|
|
JSON Output:
|
|
"""
|
|
)
|
|
return prompt
|