generated: '2026-08-14'
method: searched
source: https://github.com/evrimai/python-client/blob/main/README.md#retries
limit_count: 0
limits: []
note: >-
  NO NUMERIC RATE LIMIT IS PUBLISHED. Evrim documents no quota, window, burst or
  per-key ceiling anywhere a member of the public can reach: the published
  OpenAPI (v0.5.18, 91 operations) declares no 429 response and no rate-limit
  headers, the marketing site has no pricing or limits page, and the developer
  documentation host (docs.evrim.ai) is unreachable — its HTTPS listener fails
  the TLS handshake and plain HTTP returns Cloudflare error 1001. What IS
  established, from Evrim's own first-party Stainless-generated Python SDK, is
  that the API enforces limits at runtime and signals them the standard way: a
  429 maps to `evrim.RateLimitError` and the client backs off. The exhaustion
  status and the retry signalling below are real; the numbers are simply not
  disclosed. Recorded as limit_count 0 rather than omitted.
exhaustion:
  status: 429
  error_type: RateLimitError
  envelope: '{"detail": "..."}'
  source: sdk-error-table
response_headers:
  documented_by_provider: false
  honoured_by_first_party_sdk:
  - name: retry-after-ms
    meaning: >-
      Non-standard millisecond precision back-off; the SDK reads this FIRST,
      before `retry-after`.
  - name: retry-after
    meaning: >-
      Seconds (non-integer floats tolerated) or an HTTP-date. The SDK only honours
      a parsed value in the range 0 < n <= 60; anything larger falls back to its
      own exponential backoff.
  - name: x-should-retry
    meaning: >-
      Explicit server override — `true` forces a retry, `false` suppresses one,
      regardless of status code.
  note: >-
    These are the headers Evrim's own SDK parses off Evrim responses
    (src/evrim/_base_client.py::_parse_retry_after_header and the
    x-should-retry check), which is provider-published evidence that the API emits
    them. No X-RateLimit-* / RateLimit-* quota-disclosure header is read or
    documented, so a client cannot see remaining quota — only that it has run out.
client_retry_policy:
  max_retries_default: 2
  backoff: short exponential
  retried_conditions:
  - connection errors
  - 408 Request Timeout
  - 409 Conflict
  - 429 Too Many Requests
  - '>= 500'
  request_timeout_default_seconds: 60
  configurable: max_retries / timeout, per-client or per-request
  source: https://github.com/evrimai/python-client/blob/main/README.md#retries
evidence:
- url: https://github.com/evrimai/python-client/blob/main/README.md
  status: 200
- url: https://raw.githubusercontent.com/evrimai/python-client/main/src/evrim/_base_client.py
  status: 200
- url: https://docs.evrim.ai/
  status: 000
  note: TLS handshake failure (alert 40, no peer certificate) — docs unreachable
- url: http://docs.evrim.ai/
  status: 409
  note: Cloudflare error 1001
checked: '2026-08-14'
cross_links:
  errors: errors/evrim-problem-types.yml
  conventions: conventions/evrim-conventions.yml