Somobai平台企业服务协议
生效日期：2026-08-21
本《企业服务协议》（以下简称 “本协议”）由厦门闪企科技有限公司（以下简称 “我方”）与申请使用Somobai平台 API 服务的企业组织机构客户（以下简称 “客户”）订立。
重要提示：Somobai平台仅向合法存续的企业、组织机构提供 API 接口服务，不向社会普通自然人个人开放注册及使用本平台服务。任何自然人不得直接注册、使用本平台服务。客户通过商务对接完成企业资质核验、开通租户账号、获取 API Key，即表示客户已完整阅读并同意全部本协议条款。
第一条 服务内容
1.1 我方运营Somobai平台，向客户提供大模型 API 令牌调度、请求路由分发、算力负载均衡、限流熔断、Token 配额管控等后端技术服务；客户通过 API 接口调用Somobai平台提供的大模型相关能力。
1.2 本服务为后端 API 技术能力，客户自行负责自身业务系统、业务内容的合规性；我方调度算法仅做请求转发、令牌配额校验，不生成、不审核客户提交的业务提示词、业务文本内容。
1.3 客户需完成企业资质核验，经我方审核通过后开通租户账号、API 访问密钥、令牌配额；客户可在企业管理中心自主管理密钥、IP 白名单、子账户、令牌包额度。
第二条 令牌、配额、限流与熔断规则
2.1 令牌（Token）配额：客户购买 / 分配得到对应 Token 令牌额度，每一次 API 调用将按照计费规则扣减客户名下令牌配额；当客户令牌配额不足时，平台将拒绝新的调用请求并返回额度不足提示。
2.2 负载调度规则：平台调度算法实时监测集群 CPU、内存、接口 QPS 等负载指标，根据集群负载状态、客户令牌配额自动完成请求分发。
2.3 限流、过载熔断机制：当整体集群负载超过预设安全阈值、或客户调用频率超出约定上限，平台将自动执行限流、排队或者拒绝新请求的熔断保护策略，保障整体服务稳定性。
2.4 运维人员具备人工干预入口，可按需调整调度阈值、限流规则，干预操作全程留痕审计。
2.5 客户可在企业管理中心查看令牌消耗流水、调用记录；调度、扣减流水日志按法规要求留存不少于 6 个月。
第三条 账号与客户义务
3.1 客户应当妥善保管 API Key、访问密钥，对自身账号下所有 API 调用行为承担全部法律责任；不得泄露、转借、转售平台访问密钥。
3.2 客户保证接入平台开展的业务合法合规，不得利用本 API 服务从事违法违规业务；若客户发生高频恶意调用、攻击平台等行为，我方有权限流、暂停或者终止服务。
3.3 客户内部员工仅作为企业经办人员使用已开通的租户能力，内部工作人员不视作平台独立用户。
第四条 数据处理规则
4.1 我方调度服务运行过程中，不采集、不存储、不解析客户提交的业务提示词、业务内容文本；不主动采集任何自然人个人敏感信息（身份证、手机号等）。
4.2 平台仅采集业务运行必要指标：企业租户标识、令牌配额、请求元数据、服务器集群负载、请求时间戳等业务资源类统计数据，用于令牌校验、调度决策、故障排查、运维审计。
4.3 上述业务资源数据仅用于Somobai平台自身服务运行，不会向任何第三方对外共享、转让。数据存储于中国境内福建省厦门市云服务器，全部运算存储均在中国境内，不存在数据跨境传输。
4.4 我方按照法律要求留存调度审计日志不少于 6 个月。
第五条 客户申诉、投诉渠道
5.1 如客户对于令牌扣减结果、调度限流熔断结果、服务异常存在异议，可通过以下渠道提交申诉：
平台租户后台工单系统；
商务对接邮箱：bd@somobai.com
5.2 我方收到企业机构客户申诉后，5 个工作日内完成核查并向客户反馈处理结果，申诉处置记录归档留存。
5.3 客户注销企业租户账户、关闭令牌调用权限，可通过工单提交申请；账户关闭后将不再调用本平台调度算法。
第六条 知识产权
6.1 Somobai平台全部软件、调度算法、接口、文档知识产权归厦门闪企科技有限公司所有。客户仅获得 API 调用的非独占、不可转授权的使用权。
6.2 客户通过 API 生成的业务输出内容知识产权归属客户与上游模型服务方按约定执行，我方不对客户业务输出内容享有知识产权。
第七条 服务中断、变更与终止
7.1 我方将尽力保障服务可用性；因服务器维护、故障、网络攻击、上游模型通道异常，可能出现短暂服务波动。
7.2 发生客户严重违约、违规调用的，我方有权通知后暂停、终止 API 访问权限。
7.3 我方可以对本协议进行版本更新，更新后的协议将发布于本页面；继续使用服务视为接受更新后协议，客户不接受应当停止调用服务。
第八条 免责与责任限制
8.1 在法律允许最大范围内，我方不对客户的业务收益、间接损失承担赔偿责任；因客户自身业务、密钥泄露、客户业务内容违规带来的全部责任，均由客户自行承担。
8.2 我方仅提供后端调度转发能力，不对下游大模型生成内容承担内容审核责任。
第九条 争议解决
9.1 本协议适用中华人民共和国大陆地区法律。
9.2 双方发生争议优先协商；协商不成，任何一方可向厦门闪企科技有限公司住所地有管辖权人民法院提起诉讼。
第十条 其他
10.1 本协议为平台对外公示版本；如客户签署线下商务合同，线下合同优先级高于网页版协议。
Somobai Platform Enterprise Service Agreement
Effective date: 2026-08-21
This Enterprise Service Agreement (this "Agreement") is entered into by and between Xiamen Shanqi Technology Co., Ltd. ("we" or "us") and the enterprise or organization applying to use the Somobai platform API service (the "Customer").
Important: The Somobai platform provides API services only to legally existing enterprises and organizations. Registration and use are not open to individual members of the public, and no natural person may register for or use the platform directly. By completing enterprise verification through our business team, opening a tenant account and obtaining an API key, the Customer confirms that it has read and agrees to all terms of this Agreement.
Article 1. Services
1.1 We operate the Somobai platform and provide the Customer with back-end technical services including large-model API token scheduling, request routing, compute load balancing, rate limiting and circuit breaking, and token quota management. The Customer accesses the large-model capabilities offered by the Somobai platform through its API.
1.2 The service is a back-end API capability. The Customer is solely responsible for the compliance of its own business systems and content. Our scheduling algorithm only forwards requests and validates token quotas; it does not generate or review the prompts or text content submitted by the Customer.
1.3 The Customer must complete enterprise verification. Once approved, we will provision a tenant account, API access keys and a token quota. The Customer may manage its keys, IP allowlist, sub-accounts and token pack balances in the enterprise console.
Article 2. Tokens, Quotas, Rate Limiting and Circuit Breaking
2.1 Token quota: The Customer obtains a token quota by purchase or allocation. Each API call deducts from the Customer's token quota according to the applicable billing rules. If the quota is insufficient, the platform will reject new calls and return an insufficient-quota response.
2.2 Load scheduling: The platform's scheduling algorithm monitors load metrics such as cluster CPU, memory and API QPS in real time, and distributes requests automatically based on cluster load and the Customer's token quota.
2.3 Rate limiting and overload protection: If overall cluster load exceeds preset safety thresholds, or the Customer's call rate exceeds the agreed limit, the platform will automatically apply rate limiting, queuing or rejection of new requests to protect overall service stability.
2.4 Operations staff have a manual override to adjust scheduling thresholds and rate-limiting rules as needed; all such interventions are logged for audit.
2.5 The Customer can view token consumption and call records in the enterprise console. Scheduling and deduction logs are retained for no less than six (6) months as required by law.
Article 3. Accounts and Customer Obligations
3.1 The Customer shall keep its API keys and access credentials secure and bears full legal responsibility for all API calls made under its account. The Customer shall not disclose, lend or resell its platform access keys.
3.2 The Customer warrants that its business conducted through the platform is lawful and shall not use the API service for any illegal purpose. If the Customer engages in high-frequency malicious calls, attacks on the platform or similar conduct, we may rate-limit, suspend or terminate the service.
3.3 The Customer's employees use the provisioned tenant capabilities solely as authorized representatives of the Customer and are not treated as independent users of the platform.
Article 4. Data Processing
4.1 In operating the scheduling service, we do not collect, store or parse the prompts or text content submitted by the Customer, and we do not proactively collect any sensitive personal information of natural persons (such as ID numbers or phone numbers).
4.2 The platform collects only the operational metrics necessary to run the service: tenant identifiers, token quotas, request metadata, server cluster load, request timestamps and similar resource statistics, which are used for token validation, scheduling decisions, troubleshooting and operational audit.
4.3 The above resource data is used solely to operate the Somobai platform and will not be shared with or transferred to any third party. Data is stored on cloud servers located in Xiamen, Fujian Province, within mainland China; all computation and storage take place within China, and no cross-border data transfer occurs.
4.4 We retain scheduling audit logs for no less than six (6) months as required by law.
Article 5. Customer Disputes and Complaints
5.1 If the Customer disagrees with a token deduction, a scheduling, rate-limiting or circuit-breaking outcome, or a service incident, it may file a dispute through:
the ticketing system in the tenant console; or
our business contact email: bd@somobai.com
5.2 We will investigate and respond to the Customer within five (5) business days of receiving a dispute, and will archive the record of its handling.
5.3 To close its tenant account or revoke token access, the Customer may submit a request by ticket. Once the account is closed, the platform's scheduling algorithm will no longer be invoked on its behalf.
Article 6. Intellectual Property
6.1 All intellectual property in the Somobai platform software, scheduling algorithms, interfaces and documentation belongs to Xiamen Shanqi Technology Co., Ltd. The Customer receives only a non-exclusive, non-sublicensable right to call the API.
6.2 Ownership of intellectual property in output generated by the Customer through the API is governed by the arrangements between the Customer and the upstream model provider. We claim no intellectual property in the Customer's output.
Article 7. Service Interruption, Changes and Termination
7.1 We will use reasonable efforts to keep the service available. Server maintenance, failures, network attacks or problems with upstream model channels may cause temporary service fluctuations.
7.2 In the event of a material breach or non-compliant use by the Customer, we may suspend or terminate API access upon notice.
7.3 We may update this Agreement from time to time; updated versions will be published on this page. Continued use of the service constitutes acceptance of the updated Agreement. If the Customer does not accept the update, it should stop using the service.
Article 8. Disclaimers and Limitation of Liability
8.1 To the maximum extent permitted by law, we are not liable for the Customer's lost business profits or any indirect losses. The Customer bears full responsibility for any consequences arising from its own business, leaked keys or non-compliant business content.
8.2 We provide only back-end scheduling and forwarding, and are not responsible for content moderation of output generated by downstream large models.
Article 9. Dispute Resolution
9.1 This Agreement is governed by the laws of the mainland of the People's Republic of China.
9.2 The parties shall first seek to resolve any dispute through negotiation. If negotiation fails, either party may bring a lawsuit before the competent people's court at the domicile of Xiamen Shanqi Technology Co., Ltd.
Article 10. Miscellaneous
10.1 This Agreement is the version published by the platform. If the Customer signs a separate offline commercial contract, the offline contract prevails over this web version.