Number of automatic retries when the LLM times out
mid-reasoning during chat. Set to 0 to disable
auto-retry.
LLM Max Retries
Maximum retry attempts on transient LLM errors (timeout,
5xx). Uses exponential backoff (2s, 4s, 8s…). Set to 0
to disable.
Max Concurrent Turns per Agent
Maximum parallel queries per agent (from different
sessions). 0 = unlimited.
Max Concurrent Turns per Model
Global default for all models (when a model's own limit
is 0). 0 = unlimited.
Max Concurrent LLM (Global)
Hard cap on simultaneous LLM calls across all
agents and models. 1 = serialised (default, safest for
shared API keys). Increase only if your provider allows
concurrent calls.
Agent Queue Workers
Number of worker threads processing agent messages.
Increase takes effect immediately; decrease requires
restart.
Public History
Allow unauthenticated users to view the history and
detail pages (read-only).