Skip to main content
Open the call logs page and locate a request by time, model, status, or other filters currently available in the console. Call logs are primarily used to review billing, routing, error categories, and request status. LaoZhang API does not store API prompts or responses by default. Based on the LaoZhang API Data Policy, last verified October 5, 2026.

​
What logs may contain

Under the Data Policy, request metadata may include:
  • Account or API key identifier
  • Timestamp, model, endpoint, and route
  • Token, usage amount, or billing unit
  • Request status and error category
  • Latency and routing information
  • Information required for security, abuse prevention, and service integrity

​
What is not stored by default

LaoZhang API does not store API prompts or responses by default. Upstream model and infrastructure providers may apply their own logging, retention, and safety-review rules. Those upstream boundaries cannot be inferred from the LaoZhang API call log alone.

​
Read the charges in a log entry

Each billing entry lists the model, input (prompt) tokens, output (completion) tokens, the charge, and a line of billing details. The ratios in that line convert to prices like this: For example, a model ratio of 1, group ratio of 1, and completion ratio of 5 mean $2 input and $10 output per 1M tokens, which matches gpt-6-sol in the model catalog.
  • Usage-based models: the token counts in the response’s usage come from the same source as the log, so you can check the charge yourself. Cost = (uncached input × input price + cached input × cache-read price + output × output price) ÷ 1,000,000 × group ratio.
  • Reasoning models: reasoning tokens count as output even though they don’t appear in the reply.
  • Per-call models: the log may show token counts, but the charge doesn’t come from them; identical calls cost the same each time.
  • The API response doesn’t include a charge; use usage and the prices to compute cost.
Failed requests that weren’t charged don’t appear as billing entries. Requests rejected before they run, such as for bad parameters, failed authentication, or a balance that can’t cover the pre-authorized charge, return their error only in the API response. Conversely, a billing entry means the request reached the upstream provider and used resources; if your client got no result, the problem happened on the way back.

​
Retention

Troubleshooting logs are generally retained for 7 days. Relevant metadata may be retained longer where reasonably necessary for security, abuse investigation, billing disputes, legal compliance, or service integrity. Billing records, transaction records, account and security events, and support materials follow separate retention purposes described in the Data Policy.

​
Troubleshooting a failed call

  1. Record the exact failure time and time zone.
  2. Confirm the model ID, endpoint, and API key group.
  3. Review the HTTP status, platform error, and whether usage was charged.
  4. Reproduce with a minimal request to separate parameter, permission, balance, route, and upstream errors.
  5. If unresolved, send redacted evidence to support.
Provide:
  • Account email or username
  • Request time and time zone
  • Model ID, endpoint, and token group
  • HTTP status, error category, and redacted error text
  • A console request identifier that support can use for investigation
  • Retry attempts and results
Do not send complete API keys, passwords, full prompts or responses, regulated data, or unrelated customer information. Do not assume that a public call-log API exists unless it is documented in the current official API reference or console.