HOARDED FORTUNE GROUP PTE. LTD. / 新加坡鸿德集团 Production inference 生产级推理方案 REQUIREMENTS CHECKLIST / 双语需求清单 Project requirements record. Signed quotations and service agreements define pricing, capacity and delivery terms. 项目需求记录。签署的报价与服务协议规定价格、容量和交付条件。 Request profile / 请求特征 Provide model/precision, input and output token distributions, context length, arrival pattern, concurrency and streaming requirements. 提供模型及精度、输入输出 Token 分布、上下文长度、请求到达模式、并发及流式要求。 Customer input / 客户填写：________________ Service objective / 服务目标 Define latency percentiles, useful throughput, availability measurement and acceptable error behaviour. Distinguish interactive traffic from offline batch jobs. 明确延迟分位数、有效吞吐、可用性统计及可接受错误行为，区分交互流量与离线批处理。 Customer input / 客户填写：________________ Serving configuration / 服务配置 Runtime revision; model replicas; concurrency limits; cache policy. 运行时版本、模型副本、并发限制、缓存策略。 Required scope / 所需范围：________________ Traffic and access / 流量与访问 Endpoint boundary; identity; traffic limits; network ownership. 端点边界、身份、流量限制、网络责任。 Required scope / 所需范围：________________ Continuity and observability / 连续性与可观测性 Failure response; capacity trigger; dashboards; incident ownership. 故障响应、容量触发条件、仪表板、事件责任。 Required scope / 所需范围：________________ Latency distribution / 延迟分布 Measure time to first token and inter-token latency for streaming workloads, and end-to-end latency percentiles under stated load. 流式负载测量首个 Token 及 Token 间延迟，并在明确负载下记录端到端延迟分位数。 Agreed target and conditions / 约定目标及条件：________________ Throughput and concurrency / 吞吐与并发 Report completed requests or output tokens per second alongside concurrency, input/output profile, errors and latency constraints. 每秒完成请求或输出 Token 数需同时说明并发、输入输出特征、错误及延迟约束。 Agreed target and conditions / 约定目标及条件：________________ Recovery and quality / 恢复与质量 Test overload, replica failure and restart behaviour. Validate accuracy using an agreed evaluation set. 测试过载、副本故障及重启行为，并以约定评估集验证准确性。 Agreed target and conditions / 约定目标及条件：________________ Additional review / 补充审阅 Region and access locations / 区域及访问地点：________________ Contracting entity and approved contact / 签约主体及认可联系：________________ Delivery window and project gates / 交付窗口及阶段条件：________________ Currency, billing unit, term and budget basis / 币种、计费单位、期限及预算口径：________________ Support and responsibility split / 支持及责任分工：________________ Data retention, export and deletion / 数据保留、导出及删除：________________ Acceptance owner and evidence format / 验收负责人及证据格式：________________ Credentials, identity documents and sensitive datasets use designated secure channels. 凭据、身份证明文件及敏感数据集通过指定安全渠道交付。