Terms of service
Effective date:
Service status: the direct API and RapidAPI connection have passed synthetic integration tests. Marketplace availability, subscription prices and quotas are shown in the RapidAPI listing. OpenRouter acceptance remains unverified.
These service terms describe the Myworkingmemory Specialist Reranking API, operated by Myworkingmemory LLC. The direct API and RapidAPI connection have passed synthetic integration tests. Marketplace availability, prices and quotas are shown in the RapidAPI listing. OpenRouter acceptance remains unverified.
Access and purpose
Use the API only through an authorization key provided for your access. Keep the key private and do not place it in a URL, public code or a support message. Send only material you are authorized to provide for processing.
The service compares a query with candidate documents and returns machine-generated relevance scores and a ranked result. Scores are estimates of relevance, not a determination that a document is true, complete or appropriate for a particular decision. Review results before relying on them.
Your content
You retain your rights in the queries and documents you supply. The service requests no assignment of your ownership. You authorize only the processing needed to handle the request and return its result, including the network delivery needed for that request.
Our application does not use request content for model training or deliberately retain it as a document collection, content log or reusable response cache. The privacy policy explains the data handling and the limits of memory-erasure and infrastructure claims.
Request limits
The initial gateway accepts no more than 32 text documents and 64 KiB of JSON in one request. Each query-and-document pair is limited to 7,168 UTF-8 bytes, including the query, within an 8,192-token backend context configuration. These are distinct limits: the gateway uses a conservative byte check and reserves space for the model's scoring prompt.
Inputs outside the accepted format or limits are declined rather than silently shortened by the gateway. The initial configuration handles one inference request at a time and sets a 60-second backend timeout. A busy request can receive a retry response. The service supports text reranking, not general chat generation.
Availability and changes
Availability depends on the serving Mac, model process, network and external routing systems. The initial service has no stated service-level agreement or guarantee of uninterrupted availability, particular score accuracy, speed or capacity.
Model and routing acceptance must be verified before the service is represented as production-ready. Operational limits and supported features may change as the service is tested. Published service information will identify material changes to its operation or data handling.
Pricing and platform billing
These terms set no customer price and create no payment commitment. When access is offered through a marketplace, the applicable displayed offering and platform billing process must identify the price before paid use. Myworkingmemory's provider application, preparation work and a working local model are not evidence that OpenRouter has accepted the service.
Contact
For access, operation or service questions, contact dean@myworkingmemory.ai. Please do not include confidential documents or access credentials.