From 590d39a0250f3b213e49090cbd02e22acd8cf615 Mon Sep 17 00:00:00 2001
From: DragonAssassin-one <2819637808@qq.com>
Date: Mon, 11 May 2026 10:51:46 +0800
Subject: [PATCH] feat(docs): Add proxy configuration settings and detailed
 router settings reference

---
 docs/my-website/docs/proxy/config_settings.md | 72 +++++++++++++++++++
 1 file changed, 72 insertions(+)
 create mode 100644 docs/my-website/docs/proxy/config_settings.md

diff --git a/docs/my-website/docs/proxy/config_settings.md b/docs/my-website/docs/proxy/config_settings.md
new file mode 100644
index 00000000000..70d830fbf96
--- /dev/null
+++ b/docs/my-website/docs/proxy/config_settings.md
@@ -0,0 +1,72 @@
+# Proxy config settings
+#
+# Note: LiteLLM's public docs are maintained in a separate repository.
+# This file exists to keep in-repo reference checks in sync with Router settings.
+
+## Router
+
+### router_settings - Reference
+
+| key | description
+|---|---
+| alerting_config | Router alerting configuration.
+| allowed_fails | Number of times a deployment can fail before being added to cooldown.
+| allowed_fails_policy | Allowed fails policy configuration.
+| assistants_config | Assistants/Responses-related configuration.
+| cache_kwargs | Additional kwargs to pass to the Router cache backend.
+| cache_responses | Enable caching of responses.
+| caching_groups | Configure caching across model groups.
+| client_ttl | Time-to-live for cached clients (seconds).
+| content_policy_fallbacks | Fallback behavior for content policy errors.
+| context_window_fallbacks | Fallback behavior for context window limit errors.
+| cooldown_time | Cooldown time (seconds) after a deployment failure.
+| debug_level | Debug level for Router logging.
+| default_fallbacks | Default fallback models/deployments.
+| default_litellm_params | Default parameters applied to completion calls.
+| default_max_parallel_requests | Default max parallel requests (scheduler).
+| default_priority | Default request priority (scheduler).
+| deployment_affinity_ttl_seconds | TTL (seconds) for user-key -> deployment affinity mapping.
+| disable_cooldowns | Disable deployment cooldown behavior.
+| enable_health_check_routing | Enable health-check-aware routing.
+| enable_pre_call_checks | Enable pre-call checks filtering deployments (e.g. context window checks).
+| enable_tag_filtering | Enable tag-based deployment filtering.
+| fallbacks | Model fallback configuration.
+| guardrail_list | Guardrails configuration list.
+| health_check_ignore_transient_errors | Ignore transient errors in health checks.
+| health_check_staleness_threshold | Threshold (seconds) for considering health checks stale.
+| ignore_invalid_deployments | Ignore invalid deployments instead of raising errors.
+| max_fallbacks | Maximum number of fallbacks to attempt.
+| model_group_affinity_config | Model group affinity configuration.
+| model_group_alias | Alias mapping for model groups.
+| model_group_retry_policy | Retry policy overrides per model group.
+| model_list | List of model deployments.
+| num_retries | Number of retries for failed requests.
+| optional_pre_call_checks | Optional pre-call checks configuration.
+| polling_interval | Polling interval for scheduler completion (seconds).
+| prompt_prefix_affinity_min_tokens | Minimum canonical prompt token count before prompt-prefix affinity applies (canonical JSON tokenization, not raw prompt tokens).
+| prompt_prefix_affinity_tokens | Number of canonical prompt-prefix tokens used for deterministic prompt-prefix routing (canonical JSON tokenization, not raw prompt tokens).
+| provider_budget_config | Provider budget configuration.
+| redis_db | Redis DB index.
+| redis_host | Redis host.
+| redis_password | Redis password.
+| redis_port | Redis port.
+| redis_url | Redis URL.
+| retry_after | Minimum time to wait before retrying a failed request (seconds).
+| retry_policy | Retry policy configuration.
+| router_general_settings | Router general settings object.
+| routing_groups | Routing groups configuration.
+| routing_strategy | Routing strategy used for the implicit default group.
+| routing_strategy_args | Arguments for the routing strategy (e.g. latency window).
+| search_tools | Search tools configuration.
+| set_verbose | Enable verbose logging.
+| stream_timeout | Stream timeout (seconds).
+| tag_filtering_match_any | Tag filtering mode (match any vs match all).
+| timeout | Request timeout (seconds).
+
+### router_settings - Notes
+
+`prompt_prefix_affinity_tokens` and `prompt_prefix_affinity_min_tokens` are measured
+against tokenization of the canonical JSON representation constructed for the request
+(field names, separators, quotes, etc.), not the provider-specific "raw prompt token"
+count after rendering the final prompt.
+