Updated test cases to use the new EvalSet schema to store test data. Also, added a utility to help migrate existing tests files to the new schema.

Also, migrated existing test files to the new schema and deleted test session files as they are no longer needed.

PiperOrigin-RevId: 759318735
This commit is contained in:
Ankur Sharma
2025-05-15 15:09:30 -07:00
committed by Copybara-Service
parent a71d9ea9a1
commit 1c23556225
17 changed files with 1110 additions and 231 deletions

View File

@@ -1,5 +1,65 @@
[{
"query": "Turn off device_2 in the Bedroom.",
"expected_tool_use": [{"tool_name": "set_device_info", "tool_input": {"location": "Bedroom", "device_id": "device_2", "status": "OFF"}}],
"reference": "I have set the device_2 status to off."
}]
{
"eval_set_id": "b305bd06-38c5-4796-b9c7-d9c7454338b9",
"name": "b305bd06-38c5-4796-b9c7-d9c7454338b9",
"description": null,
"eval_cases": [
{
"eval_id": "tests/integration/fixture/home_automation_agent/simple_test.test.json",
"conversation": [
{
"invocation_id": "b7982664-0ab6-47cc-ab13-326656afdf75",
"user_content": {
"parts": [
{
"video_metadata": null,
"thought": null,
"code_execution_result": null,
"executable_code": null,
"file_data": null,
"function_call": null,
"function_response": null,
"inline_data": null,
"text": "Turn off device_2 in the Bedroom."
}
],
"role": "user"
},
"final_response": {
"parts": [
{
"video_metadata": null,
"thought": null,
"code_execution_result": null,
"executable_code": null,
"file_data": null,
"function_call": null,
"function_response": null,
"inline_data": null,
"text": "I have set the device_2 status to off."
}
],
"role": "model"
},
"intermediate_data": {
"tool_uses": [
{
"id": null,
"args": {
"location": "Bedroom",
"device_id": "device_2",
"status": "OFF"
},
"name": "set_device_info"
}
],
"intermediate_responses": []
},
"creation_timestamp": 1747337309.2360144
}
],
"session_input": null,
"creation_timestamp": 1747337309.2360282
}
],
"creation_timestamp": 1747337309.2360387
}