rubric_based_tool_use_quality_v1
https://google.github.io/adk-docs/evaluate/criteria/#rubric_based_final_response_quality_v1
#ADK_Evaluation_Criteria
https://github.com/google/adk-python/blob/v1.21.0/src/google/adk/evaluation/rubric_based_final_response_quality_v1.py#L235
The evaluator uses a set of rubrics to assess the quality of the agent's final response.
For each rubric, this evaluator will generate a confidence score between 0 and 1, where 0 means that agent's response did not satisfy the rubric at all and 1 means complete adherence. Value closer to 1 are desirable.
_RUBRIC_BASED_FINAL_RESPONSE_QUALITY_V1_PROMPT
SPECIAL INSTRUCTION: think silently. Silent thinking token budget: 10240 tokens.
Mission
Rubric
Key Evaluation Principles
Output Format
REMEMBER: Your answer will help improve the AI agent. It is important to determine the fulfillment of the properties correctly. Even answering "no" will improve the agent! Respond in pure text, not json.
Example
Input
<user_prompt>
developer_instructions
tool_declarations
user_input
<response>
response_steps
<properties>
rubrics
Output
Your Turn