Sessions

Routing affinity

Text caching is automatic. An optional prompt_cache_key improves reuse within a workspace. Cache lifetime, minimum size and eviction are not published. See Text caching.

session_id helps choose the first model server within your workspace. When it is absent, prompt_cache_key supplies the routing hint.

[ Choose routing hint ]
  |
  |-- session_id set
  |   use session_id
  |
  |-- else cache key set
  |   use prompt_cache_key
  |
  '-- neither set
      no explicit hint

[ Hint is workspace scoped ]
[ Send a complete request ]

The hint does not guarantee a particular server or cached input. Each decide request still supplies its state and questions.

Request state

Setting session_id on a decide request does not retain state.

Send each request to the stateless decide endpoint.

Route refusal

Listed API

404 long_context_not_available: This route is not available.

Retry

false. There is no Retry-After header.

These refusals apply to /v1/system-zero/sessions and its operation paths.