revert os flag thing

components/src/dynamo/frontend/main.py

@@ -151,12 +151,6 @@ def parse_args():
         default=float(os.environ.get("DYN_ROUTER_PRUNE_TARGET_RATIO", "0.8")),
         help="KV Router: Target size ratio after pruning when KV events are disabled. Only used when --no-kv-events is set. Can be set via DYN_ROUTER_PRUNE_TARGET_RATIO env var (default: 0.8).",
     )
-    parser.add_argument(
-        "--enable-local-kvindexers",  # TODO (karen-sy) eventually will be equivalent to a "bypass Jetstream" option
-        action="store_true",
-        default=False,
-        help="KV Router: Enable worker-local kvindexers for worker-specific event storage (default disabled).",
-    )
     parser.add_argument(
         "--namespace",
         type=str,
@@ -298,13 +292,6 @@ def signal_handler():
 
     if flags.router_mode == "kv":
         router_mode = RouterMode.KV
-
-        if flags.enable_local_kvindexers:
-            # set env var which propagates to worker-side publisher
-            os.environ["DYN_ENABLE_LOCAL_KVINDEXERS"] = "1"
-        else:
-            os.environ.pop("DYN_ENABLE_LOCAL_KVINDEXERS", None)
-
         kv_router_config = KvRouterConfig(
             overlap_score_weight=flags.kv_overlap_score_weight,
             router_temperature=flags.router_temperature,
@@ -316,7 +303,6 @@ def signal_handler():
             router_ttl_secs=flags.router_ttl,
             router_max_tree_size=flags.router_max_tree_size,
             router_prune_target_ratio=flags.router_prune_target_ratio,
-            enable_local_kvindexers=flags.enable_local_kvindexers,
         )
     elif flags.router_mode == "random":
         router_mode = RouterMode.Random

components/src/dynamo/router/__main__.py

@@ -15,7 +15,6 @@
 import argparse
 import asyncio
 import logging
-import os
 from typing import Optional
 
 import uvloop
@@ -240,12 +239,6 @@ def parse_args():
         help="KV Router: Target size ratio after pruning (0.0-1.0). Only used when --no-kv-events is set. Determines how aggressively to prune the tree (default: 0.8)",
     )
 
-    parser.add_argument(
-        "--enable-local-kvindexers",
-        action="store_true",
-        default=False,
-        help="KV Router: Enable worker-local kvindexers for worker-specific event storage (default disabled).",
-    )
     return parser.parse_args()
 
 
@@ -264,12 +257,6 @@ async def worker(runtime: DistributedRuntime):
         )
     namespace = endpoint_parts[0]
 
-    if args.enable_local_kvindexers:
-        # set env var which propagates to worker-side publisher
-        os.environ["DYN_ENABLE_LOCAL_KVINDEXERS"] = "1"
-    else:
-        os.environ.pop("DYN_ENABLE_LOCAL_KVINDEXERS", None)
-
     logger.info("Starting Standalone Router Service")
     logger.debug(
         f"Configuration: endpoint={args.endpoint}, block_size={args.block_size}, "
@@ -281,8 +268,7 @@ async def worker(runtime: DistributedRuntime):
         f"router_track_active_blocks={args.router_track_active_blocks}, "
         f"router_ttl_secs={args.router_ttl_secs}, "
         f"router_max_tree_size={args.router_max_tree_size}, "
-        f"router_prune_target_ratio={args.router_prune_target_ratio}, "
-        f"enable_local_kvindexers={args.enable_local_kvindexers}"
+        f"router_prune_target_ratio={args.router_prune_target_ratio}"
     )
 
     # Create KvRouter configuration
@@ -297,7 +283,6 @@ async def worker(runtime: DistributedRuntime):
         router_ttl_secs=args.router_ttl_secs,
         router_max_tree_size=args.router_max_tree_size,
         router_prune_target_ratio=args.router_prune_target_ratio,
-        enable_local_kvindexers=args.enable_local_kvindexers,
     )
 
     # Create service component - use "router" as component name

launch/dynamo-run/src/flags.rs

@@ -192,7 +192,6 @@ impl Flags {
                 None,
                 None,
                 None,
-                None,
             ),
         )
     }

lib/bindings/c/src/lib.rs

@@ -460,7 +460,6 @@ pub unsafe extern "C" fn dynamo_create_worker_selection_pipeline(
                 None, // router_ttl_secs
                 None, // router_max_tree_size
                 None, // router_prune_target_ratio
-                None, // enable_local_kvindexers
             ))
         } else {
             None

lib/bindings/python/rust/llm/entrypoint.rs

@@ -41,7 +41,7 @@ impl KvRouterConfig {
 #[pymethods]
 impl KvRouterConfig {
     #[new]
-    #[pyo3(signature = (overlap_score_weight=1.0, router_temperature=0.0, use_kv_events=true, router_replica_sync=false, router_track_active_blocks=true, router_snapshot_threshold=1000000, router_reset_states=false, router_ttl_secs=120.0, router_max_tree_size=1024, router_prune_target_ratio=0.8, enable_local_kvindexers=false))]
+    #[pyo3(signature = (overlap_score_weight=1.0, router_temperature=0.0, use_kv_events=true, router_replica_sync=false, router_track_active_blocks=true, router_snapshot_threshold=1000000, router_reset_states=false, router_ttl_secs=120.0, router_max_tree_size=1024, router_prune_target_ratio=0.8))]
     #[allow(clippy::too_many_arguments)]
     fn new(
         overlap_score_weight: f64,
@@ -54,7 +54,6 @@ impl KvRouterConfig {
         router_ttl_secs: f64,
         router_max_tree_size: usize,
         router_prune_target_ratio: f64,
-        enable_local_kvindexers: bool,
     ) -> Self {
         KvRouterConfig {
             inner: RsKvRouterConfig {
@@ -68,7 +67,6 @@ impl KvRouterConfig {
                 router_ttl_secs,
                 router_max_tree_size,
                 router_prune_target_ratio,
-                enable_local_kvindexers,
             },
         }
     }

lib/bindings/python/src/dynamo/_core.pyi

@@ -1036,7 +1036,6 @@ class KvRouterConfig:
         router_ttl_secs: float = 120.0,
         router_max_tree_size: int = 1024,
         router_prune_target_ratio: float = 0.8,
-        enable_local_kvindexers: bool = False,
     ) -> None:
         """
         Create a KV router configuration.
@@ -1052,7 +1051,6 @@ class KvRouterConfig:
             router_ttl_secs: TTL for blocks in seconds when not using KV events (default: 120.0)
             router_max_tree_size: Maximum tree size before pruning (default: 1024)
             router_prune_target_ratio: Target size ratio after pruning (default: 0.8)
-            enable_local_kvindexers: Enable worker-local KV indexers for recovery queries (default: False)
         """
         ...
 

lib/llm/src/kv_router.rs

@@ -150,9 +150,6 @@ pub struct KvRouterConfig {
 
     /// Target size ratio after pruning (only used when use_kv_events is false, default: 0.8)
     pub router_prune_target_ratio: f64,
-
-    /// Whether to use worker-local kvindexers for worker-specific event storage
-    pub enable_local_kvindexers: bool,
 }
 
 impl Default for KvRouterConfig {
@@ -168,7 +165,6 @@ impl Default for KvRouterConfig {
             router_ttl_secs: 120.0,
             router_max_tree_size: 1024,
             router_prune_target_ratio: 0.8,
-            enable_local_kvindexers: false,
         }
     }
 }
@@ -188,7 +184,6 @@ impl KvRouterConfig {
         router_ttl_secs: Option<f64>,
         router_max_tree_size: Option<usize>,
         router_prune_target_ratio: Option<f64>,
-        enable_local_kvindexers: Option<bool>,
     ) -> Self {
         let default = Self::default();
         Self {
@@ -205,8 +200,6 @@ impl KvRouterConfig {
             router_max_tree_size: router_max_tree_size.unwrap_or(default.router_max_tree_size),
             router_prune_target_ratio: router_prune_target_ratio
                 .unwrap_or(default.router_prune_target_ratio),
-            enable_local_kvindexers: enable_local_kvindexers
-                .unwrap_or(default.enable_local_kvindexers),
         }
     }
 }

lib/llm/src/kv_router/publisher.rs

@@ -1,7 +1,6 @@
 // SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 // SPDX-License-Identifier: Apache-2.0
 
-use std::env;
 use std::fmt;
 use std::sync::atomic::{AtomicU32, Ordering};
 use std::sync::{Arc, OnceLock};
@@ -112,15 +111,7 @@ impl KvEventPublisher {
         kv_block_size: u32,
         source_config: Option<KvEventSourceConfig>,
     ) -> Result<Self> {
-        let enable_local_indexer = env::var("DYN_ENABLE_LOCAL_KVINDEXERS")
-            .map(|value| value == "1")
-            .unwrap_or(false);
-        Self::new_with_local_indexer(
-            component,
-            kv_block_size,
-            source_config,
-            enable_local_indexer,
-        )
+        Self::new_with_local_indexer(component, kv_block_size, source_config, true)
     }
 
     pub fn new_with_local_indexer(
@@ -2189,12 +2180,6 @@ mod test_integration_publisher {
     #[tokio::test]
     #[ignore] // Requires NATS/etcd. Run with: cargo test --package dynamo-llm --lib --features integration test_distributed_kvindexer_e2e -- --ignored --nocapture
     async fn test_distributed_kvindexer_e2e() -> anyhow::Result<()> {
-        unsafe {
-            // Typical use case would set this env var via
-            // --enable-local-kvindexers flag in frontend or standalone router
-            env::set_var("DYN_ENABLE_LOCAL_KVINDEXERS", "1");
-        }
-
         use crate::kv_router::scheduler::DefaultWorkerSelector;
         use crate::kv_router::{
             KvPushRouter, KvRouter, KvRouterConfig, worker_query::WorkerQueryClient,
@@ -2379,10 +2364,6 @@ mod test_integration_publisher {
         for handle in server_handles {
             handle.abort();
         }
-        unsafe {
-            env::remove_var("DYN_ENABLE_LOCAL_KVINDEXERS");
-        }
-
         distributed1.shutdown();
         distributed2.shutdown();
         router_distributed.shutdown();