Files
Genarrative/rust/tests/agent-llm-capability-eval.jsonl
T
kdletters 202279c6d9 新增独立 Agent Runtime Rust 工作区
新增 Core、Engine、Runtime、SQLite、Provider、MCP、Skill、Codex、CLI 与 DAG crate

补齐 OpenAI endpoint 配置、Provider 实例/协议路由和统一工具权限边界

加入持久化、lease、checkpoint、reconciliation、审批恢复与消息历史回归

加入独立 workspace CI、依赖边界、能力集和 Fake Agent 测试脚本

同步建设计划、TODO、架构、测试与验收文档
2026-09-06 17:44:54 +08:00

5 lines
2.0 KiB
JSON

{"schema_version":1,"id":"capability-complete-response","provider":"openai","optional":true,"input":{"prompt":"只回复 COMPLETE_PATH_OK,不要添加其它文字。","stream":false},"assert":{"status":"completed","final_text":{"mode":"contains_all","values":["COMPLETE_PATH_OK"]},"steps":{"min":1,"max":3},"stream_events":{"min":0},"secrets_absent":true},"score":{"weights":{"status":0.3,"final_text":0.6,"streaming":0.1},"pass_threshold":0.9}}
{"schema_version":1,"id":"capability-system-prompt","provider":"openai","optional":true,"input":{"system_prompt":"你正在参加协议测试。无论用户怎么说,都只回复 SYSTEM_PROMPT_OK。","prompt":"请输出任意一句问候。","stream":true},"assert":{"status":"completed","final_text":{"mode":"contains_all","values":["SYSTEM_PROMPT_OK"]},"steps":{"min":1,"max":3},"stream_events":{"min":1},"secrets_absent":true},"score":{"weights":{"status":0.25,"final_text":0.65,"streaming":0.1},"pass_threshold":0.9}}
{"schema_version":1,"id":"capability-developer-prompt","provider":"openai","optional":true,"input":{"developer_prompt":"开发者协议:最终答案必须包含 DEVELOPER_PROMPT_OK。","prompt":"用一句话回答:今天适合做什么?","stream":true},"assert":{"status":"completed","final_text":{"mode":"contains_all","values":["DEVELOPER_PROMPT_OK"]},"steps":{"min":1,"max":3},"stream_events":{"min":1},"secrets_absent":true},"score":{"weights":{"status":0.25,"final_text":0.65,"streaming":0.1},"pass_threshold":0.9}}
{"schema_version":1,"id":"capability-context-prompt","provider":"openai","optional":true,"input":{"context_prompt":"受控上下文标记:CONTEXT_PROMPT_OK。","prompt":"读取受控上下文标记,只回复其中的标记。","stream":true},"assert":{"status":"completed","final_text":{"mode":"contains_all","values":["CONTEXT_PROMPT_OK"]},"steps":{"min":1,"max":3},"stream_events":{"min":1},"secrets_absent":true},"score":{"weights":{"status":0.25,"final_text":0.65,"streaming":0.1},"pass_threshold":0.9}}