docs: add NIXL backend configuration and fix multiple typos (#5564)

Signed-off-by: Abhishek Kumar Gupta (AbhiOnGithub)
Signed-off-by: Abhishek Kumar Gupta <mail2abhishekgupta@gmail.com>
Co-authored-by: dagil-nvidia <dagil@nvidia.com>

components/src/dynamo/vllm/args.py

@@ -61,7 +61,7 @@ class Config:
 
     # multimodal options
     multimodal_processor: bool = False
-    # Emebdding Cache Processor is different from the regular processor
+    # Embedding Cache Processor is different from the regular processor
     # TODO: Have a single processor for all cases and adopting rust based processor
     ec_processor: bool = False
     multimodal_encode_worker: bool = False

docs/backends/trtllm/kv-cache-transfer.md

@@ -30,7 +30,39 @@ By default, TensorRT-LLM uses **NIXL** (NVIDIA Inference Xfer Library) with UCX
 
 ### Specify Backends for NIXL
 
-TODO: Add instructions for how to specify different backends for NIXL.
+NIXL supports multiple communication backends that can be configured via environment variables. By default, UCX is used if no backends are explicitly specified.
+
+**Environment Variable Format:**
+```bash
+DYN_KVBM_NIXL_BACKEND_<BACKEND>=<value>
+```
+
+**Supported Backends:**
+- `UCX` - Unified Communication X (default)
+- `GDS` - GPU Direct Storage
+
+**Examples:**
+```bash
+# Enable UCX backend (default behavior)
+export DYN_KVBM_NIXL_BACKEND_UCX=true
+
+# Enable GDS backend
+export DYN_KVBM_NIXL_BACKEND_GDS=true
+
+# Enable multiple backends
+export DYN_KVBM_NIXL_BACKEND_UCX=true
+export DYN_KVBM_NIXL_BACKEND_GDS=true
+
+# Explicitly disable a backend
+export DYN_KVBM_NIXL_BACKEND_GDS=false
+```
+
+**Valid Values:**
+- `true`, `1`, `on`, `yes` - Enable the backend
+- `false`, `0`, `off`, `no` - Disable the backend
+
+> [!Note]
+> If no `DYN_KVBM_NIXL_BACKEND_*` environment variables are set, UCX is used as the default backend.
 
 ## Alternative Method: UCX
 

docs/backends/vllm/deepseek-r1.md

@@ -5,7 +5,7 @@ SPDX-License-Identifier: Apache-2.0
 
 # Running Deepseek R1 with Wide EP
 
-Dynamo supports running Deepseek R1 with data parallel attention and wide expert parallelism. Each data parallel attention rank is a seperate dynamo component that will emit its own KV Events and Metrics. vLLM controls the expert parallelism using the flag `--enable-expert-parallel`
+Dynamo supports running Deepseek R1 with data parallel attention and wide expert parallelism. Each data parallel attention rank is a separate dynamo component that will emit its own KV Events and Metrics. vLLM controls the expert parallelism using the flag `--enable-expert-parallel`
 
 ## Instructions
 

examples/backends/sglang/slurm_jobs/scripts/vllm/benchmark_serving.py

@@ -1284,7 +1284,7 @@ if __name__ == "__main__":
         "--percentile-metrics",
         type=str,
         default="ttft,tpot,itl",
-        help="Comma-seperated list of selected metrics to report percentils. "
+        help="Comma-separated list of selected metrics to report percentiles. "
         "This argument specifies the metrics to report percentiles. "
         'Allowed metric names are "ttft", "tpot", "itl", "e2el". '
         'Default value is "ttft,tpot,itl".',
@@ -1293,7 +1293,7 @@ if __name__ == "__main__":
         "--metric-percentiles",
         type=str,
         default="99",
-        help="Comma-seperated list of percentiles for selected metrics. "
+        help="Comma-separated list of percentiles for selected metrics. "
         'To report 25-th, 50-th, and 75-th percentiles, use "25,50,75". '
         'Default value is "99". '
         'Use "--percentile-metrics" to select metrics.',

lib/llm/src/kv_router/publisher.rs

@@ -1318,7 +1318,7 @@ mod tests_startup_helpers {
         }
         assert!(no_blocks, "worker should have no blocks after removal");
 
-        // Global kvindexer should have recieved two events (create/remove)
+        // Global kvindexer should have received two events (create/remove)
         let published = published.lock().unwrap();
         assert_eq!(
             published.len(),
@@ -1397,7 +1397,7 @@ mod tests_startup_helpers {
         }
         assert!(no_blocks, "worker should have no blocks after clearing");
 
-        // Global kvindexer should have recieved two events (create/remove)
+        // Global kvindexer should have received two events (create/remove)
         let published = published.lock().unwrap();
         assert_eq!(
             published.len(),