Support
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.
The Myworkingmemory Specialist Reranking API, operated by Myworkingmemory LLC, compares a search query with candidate text documents and returns relevance-ranked results. This is a specialist document-ranking service.
Current service status
The direct API and RapidAPI connection have passed synthetic integration tests. RapidAPI marketplace availability is shown in its listing. The OpenRouter provider application has been submitted, but acceptance and availability there remain unverified.
Connecting an authorized client
The API base URL is https://deans-macbook-pro.tail5c2242.ts.net/v1. Use the private authorization key supplied for your access. This page does not publish an access key.
The model identifier is qwen/qwen3-reranker-8b. Send a JSON request to POST /v1/rerank with a text query and a documents array. Each document may be a string or an object containing only a text field. Set top_n to the number of results you want.
Use the Authorization: Bearer header for your private key and Content-Type: application/json. Do not put the key in a URL. The gateway checks the fixed model identifier and does not accept arbitrary upstream addresses.
The response contains ranked result indexes, relevance scores and the model's actual token usage. Results omit document text unless you explicitly request return_documents: true.
RapidAPI integration status
The RapidAPI connection has passed a synthetic integration test. Marketplace availability, prices and quotas are shown in the RapidAPI listing. Use the endpoint and consumer credentials provided in that listing when access is available. The private key for direct API access is not a RapidAPI consumer key.
RapidAPI handles its own platform accounts and billing. Its marketplace gateway can record full requests and responses, including document content and scores. Its documentation describes different log periods: up to seven days in the Gateway documentation and up to 30 days in Provider Analytics. Exact retention and available logging controls for this listing remain unverified. Our gateway's no-content-logging configuration does not make the RapidAPI route zero-retention. See RapidAPI coverage in the privacy policy for the source links and details.
Initial operating limits
A request may contain at most 32 documents and 64 KiB of JSON. Each query-and-document pair must fit within 7,168 UTF-8 bytes, including the query. The configured backend context is 8,192 tokens, with space reserved for its scoring prompt.
Inputs outside these limits are declined rather than silently shortened by the gateway. The initial gateway handles one inference at a time and has a 60-second backend timeout. Reduce document count or length when a request exceeds a limit.
Responses and troubleshooting
An authorization response means the supplied key is absent or invalid. A request-format response means the JSON, document format, model identifier or size limit needs correction. Do not repeatedly retry an unchanged invalid request.
A busy response means another request is running; wait for the response's retry interval before trying again. A service-unavailable response means the model or route could not complete the request. Availability depends on the serving Mac and network.
GET /health checks model readiness. GET /v1/models lists model metadata. A responding health check is separate from production acceptance and marketplace availability.
Private support
Email dean@myworkingmemory.ai. Include the response status, a short description of what you expected and a small invented example that reproduces the problem. Do not include your API key, password, confidential query or private documents.
Read the privacy policy for the no-content-storage configuration and its limits. Read the terms of service for access and operating conditions.