chore: rename dynamo (#44)

Co-authored-by: Biswa Panda <biswa.panda@gmail.com>

examples/python_rs/llm/tensorrt_llm/monolith/worker.py

@@ -32,7 +32,7 @@ from tensorrt_llm.serve.openai_protocol import (
     CompletionStreamResponse,
 )
 
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 logger.set_level("debug")
 
@@ -45,7 +45,7 @@ class TensorrtLLMEngine(BaseTensorrtLLMEngine):
     def __init__(self, engine_config: LLMAPIConfig):
         super().__init__(engine_config)
 
-    @dynemo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
+    @dynamo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
     async def generate_chat(self, request):
         if self._llm_engine is None:
             raise RuntimeError("Engine not initialized")
@@ -93,7 +93,7 @@ class TensorrtLLMEngine(BaseTensorrtLLMEngine):
         except Exception as e:
             raise RuntimeError("Failed to generate: " + str(e))
 
-    @dynemo_endpoint(CompletionRequest, CompletionStreamResponse)
+    @dynamo_endpoint(CompletionRequest, CompletionStreamResponse)
     async def generate_completion(self, request):
         if self._llm_engine is None:
             raise RuntimeError("Engine not initialized")
@@ -140,13 +140,13 @@ class TensorrtLLMEngine(BaseTensorrtLLMEngine):
             raise RuntimeError("Failed to generate: " + str(e))
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_config: LLMAPIConfig):
     """
     Instantiate a `backend` component and serve the `generate` endpoint
     A `Component` can serve multiple endpoints
     """
-    component = runtime.namespace("dynemo").component("tensorrt-llm")
+    component = runtime.namespace("dynamo").component("tensorrt-llm")
     await component.create_service()
 
     completions_endpoint = component.endpoint("completions")

examples/python_rs/llm/vllm/README.md

@@ -63,7 +63,7 @@ By default the server will run on port 8080.
 
 Add model to the server:
 ```bash
-llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Llama-8B dynemo.vllm.generate
+llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Llama-8B dynamo.vllm.generate
 ```
 
 ##### Example Output
@@ -71,7 +71,7 @@ llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Llama-8B dynemo.vllm
 +------------+------------------------------------------+-----------+-----------+----------+
 | MODEL TYPE | MODEL NAME                               | NAMESPACE | COMPONENT | ENDPOINT |
 +------------+------------------------------------------+-----------+-----------+----------+
-| chat       | deepseek-ai/DeepSeek-R1-Distill-Llama-8B | dynemo | vllm      | generate |
+| chat       | deepseek-ai/DeepSeek-R1-Distill-Llama-8B | dynamo | vllm      | generate |
 +------------+------------------------------------------+-----------+-----------+----------+
 ```
 
@@ -83,7 +83,7 @@ In a separate terminal run the vllm worker:
 
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 
 # Launch worker
 cd /workspace/examples/python_rs/llm/vllm
@@ -116,7 +116,7 @@ This deployment option splits the model serving across prefill and decode worker
 **Terminal 1 - Prefill Worker:**
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 
 # Launch prefill worker
 cd /workspace/examples/python_rs/llm/vllm
@@ -126,7 +126,7 @@ VLLM_WORKER_MULTIPROC_METHOD=spawn CUDA_VISIBLE_DEVICES=0 python3 -m disaggregat
     --enforce-eager \
     --tensor-parallel-size 1 \
     --kv-transfer-config \
-    '{"kv_connector":"DynemoNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2}'
+    '{"kv_connector":"DynamoNcclConnector","kv_role":"kv_producer","kv_rank":0,"kv_parallel_size":2}'
 ```
 
 ##### Example Output
@@ -142,7 +142,7 @@ INFO 03-02 05:59:47 llm_engine.py:476] init engine (profile, create kv cache, wa
 **Terminal 2 - Decode Worker:**
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 
 # Launch decode worker
 cd /workspace/examples/python_rs/llm/vllm
@@ -152,7 +152,7 @@ VLLM_WORKER_MULTIPROC_METHOD=spawn CUDA_VISIBLE_DEVICES=1,2 python3 -m disaggreg
     --enforce-eager \
     --tensor-parallel-size 2 \
     --kv-transfer-config \
-    '{"kv_connector":"DynemoNcclConnector","kv_role":"kv_consumer","kv_rank":1,"kv_parallel_size":2}'
+    '{"kv_connector":"DynamoNcclConnector","kv_role":"kv_consumer","kv_rank":1,"kv_parallel_size":2}'
 ```
 
 The disaggregated deployment utilizes separate GPUs for prefill and decode operations, allowing for optimized resource allocation and improved performance. For more details on the disaggregated deployment, please refer to the [vLLM documentation](https://docs.vllm.ai/en/latest/features/disagg_prefill.html).
@@ -254,7 +254,7 @@ tmux ls | grep 'v-' | cut -d: -f1 | xargs -I{} tmux kill-session -t {}
 **Terminal 1 - Router:**
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 
 # Launch prefill worker
 cd /workspace/examples/python_rs/llm/vllm
@@ -270,7 +270,7 @@ You can choose only the prefix strategy for now:
 **Terminal 2 - Processor:**
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 
 # Processor must take the same args as the worker
 # This is temporary until we communicate the ModelDeploymentCard over etcd
@@ -286,7 +286,7 @@ RUST_LOG=info python3 -m kv_router.processor \
 **Terminal 3 and 4 - Workers:**
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 
 # Launch Worker 1 and Worker 2 with same command
 cd /workspace/examples/python_rs/llm/vllm
@@ -304,7 +304,7 @@ Note: block-size must be 64, otherwise Router won't work (accepts only 64 tokens
 **Terminal 5 - Client:**
 Don't forget to add the model to the server:
 ```bash
-llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Llama-8B dynemo.process.chat/completions
+llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Llama-8B dynamo.process.chat/completions
 ```
 
 ```bash
@@ -351,7 +351,7 @@ Run following commands in 4 terminals:
 **Terminal 1 - vLLM Worker:**
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 cd /workspace/examples/python_rs/llm/vllm
 
 RUST_LOG=info python3 -m preprocessor.worker --model deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
@@ -361,7 +361,7 @@ RUST_LOG=info python3 -m preprocessor.worker --model deepseek-ai/DeepSeek-R1-Dis
 
 ```bash
 # Activate virtual environment
-source /opt/dynemo/venv/bin/activate
+source /opt/dynamo/venv/bin/activate
 cd /workspace/examples/python_rs/llm/vllm
 
 RUST_LOG=info python3 -m preprocessor.processor --model deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
@@ -377,7 +377,7 @@ By default the server will run on port 8080.
 
 Add model to the server:
 ```bash
-llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B dynemo.preprocessor.generate
+llmctl http add chat-models deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B dynamo.preprocessor.generate
 ```
 
 **Terminal 4 - client**

examples/python_rs/llm/vllm/common/client.py

@@ -19,12 +19,12 @@ import asyncio
 
 import uvloop
 
-from dynemo.runtime import DistributedRuntime, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_worker
 
 from .protocol import Request
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(
     runtime: DistributedRuntime,
     component: str,
@@ -36,7 +36,7 @@ async def worker(
     Instantiate a `backend` client and call the `generate` endpoint
     """
     # get endpoint
-    endpoint = runtime.namespace("dynemo").component(component).endpoint("generate")
+    endpoint = runtime.namespace("dynamo").component(component).endpoint("generate")
 
     # create client
     client = await endpoint.client()

examples/python_rs/llm/vllm/disaggregated/decode_worker.py

@@ -31,7 +31,7 @@ from vllm.entrypoints.openai.protocol import (
 )
 from vllm.logger import logger as vllm_logger
 
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 
 class VllmDecodeEngine(BaseVllmEngine, ProcessMixIn):
@@ -54,7 +54,7 @@ class VllmDecodeEngine(BaseVllmEngine, ProcessMixIn):
         self.kv_transfer_config = engine_args.create_engine_config().kv_transfer_config
         self.kv_rank = self.kv_transfer_config.kv_rank
 
-    @dynemo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
+    @dynamo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
     async def generate(self, raw_request):
         if self.engine_client is None:
             await self.initialize()
@@ -103,17 +103,17 @@ class VllmDecodeEngine(BaseVllmEngine, ProcessMixIn):
         await prefill_output
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     """
     Instantiate a `backend` component and serve the `generate` endpoint
     A `Component` can serve multiple endpoints
     """
-    component = runtime.namespace("dynemo").component("vllm")
+    component = runtime.namespace("dynamo").component("vllm")
     await component.create_service()
 
     prefill = (
-        await runtime.namespace("dynemo")
+        await runtime.namespace("dynamo")
         .component("prefill")
         .endpoint("generate")
         .client()

examples/python_rs/llm/vllm/disaggregated/prefill_worker.py

@@ -24,7 +24,7 @@ from common.protocol import PrefillRequest, PrefillResponse
 from vllm.engine.arg_utils import AsyncEngineArgs
 from vllm.logger import logger as vllm_logger
 
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 
 class VllmPrefillEngine(BaseVllmEngine):
@@ -45,7 +45,7 @@ class VllmPrefillEngine(BaseVllmEngine):
         self.kv_transfer_config = engine_args.create_engine_config().kv_transfer_config
         self.kv_rank = self.kv_transfer_config.kv_rank
 
-    @dynemo_endpoint(PrefillRequest, PrefillResponse)
+    @dynamo_endpoint(PrefillRequest, PrefillResponse)
     async def generate(self, request):
         if self.engine_client is None:
             await self.initialize()
@@ -62,13 +62,13 @@ class VllmPrefillEngine(BaseVllmEngine):
                 yield True
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     """
     Instantiate a `backend` component and serve the `generate` endpoint
     A `Component` can serve multiple endpoints
     """
-    component = runtime.namespace("dynemo").component("prefill")
+    component = runtime.namespace("dynamo").component("prefill")
     await component.create_service()
 
     async with VllmPrefillEngine(engine_args) as prefill_engine:

examples/python_rs/llm/vllm/kv_router/processor.py

@@ -34,7 +34,7 @@ from vllm.logger import logger as vllm_logger
 from vllm.outputs import RequestOutput
 from vllm.transformers_utils.tokenizer import AnyTokenizer
 
-from dynemo.runtime import Client, DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.runtime import Client, DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 
 class RequestType(Enum):
@@ -157,38 +157,38 @@ class Processor(ProcessMixIn):
                     f"Request type {request_type} not implemented"
                 )
 
-    @dynemo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
+    @dynamo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
     async def generate_chat(self, raw_request):
         async for response in self._generate(raw_request, RequestType.CHAT):
             yield response
 
-    @dynemo_endpoint(CompletionRequest, CompletionStreamResponse)
+    @dynamo_endpoint(CompletionRequest, CompletionStreamResponse)
     async def generate_completions(self, raw_request):
         async for response in self._generate(raw_request, RequestType.COMPLETION):
             yield response
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     """
     Set up clients to the router and workers.
-    Serve the dynemo.process.chat/completions endpoint.
+    Serve the dynamo.process.chat/completions endpoint.
     """
     workers_client = (
-        await runtime.namespace("dynemo")
+        await runtime.namespace("dynamo")
         .component("vllm")
         .endpoint("generate")
         .client()
     )
 
     router_client = (
-        await runtime.namespace("dynemo")
+        await runtime.namespace("dynamo")
         .component("router")
         .endpoint("generate")
         .client()
     )
 
-    preprocess_component = runtime.namespace("dynemo").component("process")
+    preprocess_component = runtime.namespace("dynamo").component("process")
     await preprocess_component.create_service()
 
     chat_endpoint = preprocess_component.endpoint("chat/completions")

examples/python_rs/llm/vllm/kv_router/router.py

@@ -23,8 +23,8 @@ import uvloop
 from common.protocol import Tokens
 from vllm.logger import logger as vllm_logger
 
-from dynemo.llm import KvIndexer, KvMetricsAggregator, KvRouter
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.llm import KvIndexer, KvMetricsAggregator, KvRouter
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 WorkerId = str
 
@@ -51,7 +51,7 @@ class Router:
         self.router = router
         self.routing_strategy = routing_strategy
 
-    @dynemo_endpoint(Tokens, WorkerId)
+    @dynamo_endpoint(Tokens, WorkerId)
     async def generate(self, request) -> AsyncIterator[WorkerId]:
         lora_id = 0
         worker_id = None
@@ -108,7 +108,7 @@ class CustomRouter:
                 )
         return current_best[0]
 
-    @dynemo_endpoint(Tokens, WorkerId)
+    @dynamo_endpoint(Tokens, WorkerId)
     async def generate(self, request) -> AsyncIterator[WorkerId]:
         lora_id = 0
         worker_id = ""
@@ -132,14 +132,14 @@ class CustomRouter:
         yield str(worker_id)
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, args: Namespace):
     """
     Set up the worker clients.
-    Serve the dynemo.router.generate endpoint.
+    Serve the dynamo.router.generate endpoint.
     """
     workers_client = (
-        await runtime.namespace("dynemo")
+        await runtime.namespace("dynamo")
         .component("vllm")
         .endpoint("generate")
         .client()
@@ -164,10 +164,10 @@ async def worker(runtime: DistributedRuntime, args: Namespace):
         + "\n".join(f"id: {id}" for id in workers_client.endpoint_ids())
     )
 
-    kv_listener = runtime.namespace("dynemo").component("vllm")
+    kv_listener = runtime.namespace("dynamo").component("vllm")
     await kv_listener.create_service()
 
-    router_component = runtime.namespace("dynemo").component("router")
+    router_component = runtime.namespace("dynamo").component("router")
     await router_component.create_service()
 
     endpoint = router_component.endpoint("generate")

examples/python_rs/llm/vllm/kv_router/worker.py

@@ -25,8 +25,8 @@ from vllm.engine.arg_utils import AsyncEngineArgs
 from vllm.logger import logger as vllm_logger
 from vllm.sampling_params import RequestOutputKind
 
-from dynemo.llm import KvMetricsPublisher
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.llm import KvMetricsPublisher
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 vllm_logger.info(f"VLLM_KV_CAPI_PATH: {os.environ['VLLM_KV_CAPI_PATH']}")
 
@@ -48,7 +48,7 @@ class VllmEngine(BaseVllmEngine):
         assert self.engine_client is not None, "engine_client was not initialized"
         self.engine_client.set_metrics_publisher(self.metrics_publisher)
 
-    @dynemo_endpoint(vLLMGenerateRequest, MyRequestOutput)
+    @dynamo_endpoint(vLLMGenerateRequest, MyRequestOutput)
     async def generate(self, request) -> AsyncIterator:
         assert (
             self.engine_client is not None
@@ -73,12 +73,12 @@ class VllmEngine(BaseVllmEngine):
             ).model_dump_json()
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     """
-    Serve the dynemo.vllm.generate endpoint.
+    Serve the dynamo.vllm.generate endpoint.
     """
-    worker_component = runtime.namespace("dynemo").component("vllm")
+    worker_component = runtime.namespace("dynamo").component("vllm")
     await worker_component.create_service()
 
     worker_endpoint = worker_component.endpoint("generate")
@@ -87,7 +87,7 @@ async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     os.environ["VLLM_WORKER_ID"] = str(VLLM_WORKER_ID)
     vllm_logger.info(f"Generate endpoint ID: {VLLM_WORKER_ID}")
 
-    VLLM_KV_NAMESPACE = "dynemo"
+    VLLM_KV_NAMESPACE = "dynamo"
     os.environ["VLLM_KV_NAMESPACE"] = str(VLLM_KV_NAMESPACE)
 
     VLLM_KV_COMPONENT = "vllm"

examples/python_rs/llm/vllm/monolith/worker.py

@@ -28,7 +28,7 @@ from vllm.entrypoints.openai.protocol import (
 )
 from vllm.logger import logger as vllm_logger
 
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 
 class VllmEngine(BaseVllmEngine, ProcessMixIn):
@@ -39,7 +39,7 @@ class VllmEngine(BaseVllmEngine, ProcessMixIn):
     def __init__(self, engine_args: AsyncEngineArgs):
         super().__init__(engine_args)
 
-    @dynemo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
+    @dynamo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
     async def generate(self, raw_request):
         if self.engine_client is None:
             await self.initialize()
@@ -71,13 +71,13 @@ class VllmEngine(BaseVllmEngine, ProcessMixIn):
             yield response
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     """
     Instantiate a `backend` component and serve the `generate` endpoint
     A `Component` can serve multiple endpoints
     """
-    component = runtime.namespace("dynemo").component("vllm")
+    component = runtime.namespace("dynamo").component("vllm")
     await component.create_service()
 
     endpoint = component.endpoint("generate")

examples/python_rs/llm/vllm/preprocessor/processor.py

@@ -19,27 +19,27 @@ import asyncio
 import uvloop
 from preprocessor.common import parse_vllm_args
 
-from dynemo.runtime import (
+from dynamo.runtime import (
     DistributedRuntime,
     ModelDeploymentCard,
     OAIChatPreprocessor,
-    dynemo_worker,
+    dynamo_worker,
 )
 
 uvloop.install()
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def preprocessor(runtime: DistributedRuntime, model_name: str, model_path: str):
     # create model deployment card
     mdc = await ModelDeploymentCard.from_local_path(model_path, model_name)
     # create preprocessor endpoint
-    component = runtime.namespace("dynemo").component("preprocessor")
+    component = runtime.namespace("dynamo").component("preprocessor")
     await component.create_service()
     endpoint = component.endpoint("generate")
 
     # create backend endpoint
-    backend = runtime.namespace("dynemo").component("backend").endpoint("generate")
+    backend = runtime.namespace("dynamo").component("backend").endpoint("generate")
 
     # start preprocessor service with next backend
     chat = OAIChatPreprocessor(mdc, endpoint, next=backend)

examples/python_rs/llm/vllm/preprocessor/worker.py

@@ -28,12 +28,12 @@ from vllm.entrypoints.openai.api_server import (
 )
 from vllm.outputs import CompletionOutput
 
-from dynemo.runtime import (
+from dynamo.runtime import (
     Backend,
     DistributedRuntime,
     ModelDeploymentCard,
-    dynemo_endpoint,
-    dynemo_worker,
+    dynamo_endpoint,
+    dynamo_worker,
 )
 
 finish_reason_map = {
@@ -107,7 +107,7 @@ class VllmEngine(AsyncContextDecorator):
         }
         return SamplingParams(**sampling_params)
 
-    @dynemo_endpoint(Any, CompletionOutput)
+    @dynamo_endpoint(Any, CompletionOutput)
     async def generate(self, request):
         state = DeltaState()
         request_id = str(uuid.uuid4())
@@ -122,13 +122,13 @@ class VllmEngine(AsyncContextDecorator):
                 yield self.to_backend_output(choice, delta_token_ids)
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: NvAsyncEngineArgs):
     """
     Instantiate a `backend` component and serve the `generate` endpoint
     A `Component` can serve multiple endpoints
     """
-    component = runtime.namespace("dynemo").component("backend")
+    component = runtime.namespace("dynamo").component("backend")
     await component.create_service()
 
     endpoint = component.endpoint("generate")

examples/python_rs/llm/vllm/scripts/disaggregated.sh

@@ -21,12 +21,12 @@
 if [ $# -gt 2 ]; then
     echo "Usage: $0 [model_name] [endpoint_name]"
     echo "Optional: model_name (default: deepseek-ai/DeepSeek-R1-Distill-Llama-8B)"
-    echo "Optional: endpoint_name (default: dynemo.vllm.generate)"
+    echo "Optional: endpoint_name (default: dynamo.vllm.generate)"
     exit 1
 fi
 
 MODEL_NAME=${1:-"deepseek-ai/DeepSeek-R1-Distill-Llama-8B"}
-ENDPOINT_NAME=${2:-"dynemo.vllm.generate"}
+ENDPOINT_NAME=${2:-"dynamo.vllm.generate"}
 SESSION_NAME="vllm_disagg"
 WORKDIR="$(dirname $0)/.."
 INIT_CMD="cd $WORKDIR"

examples/python_rs/llm/vllm/scripts/kv-router-run.sh

@@ -22,14 +22,14 @@ if [ $# -lt 2 ]; then
     echo "Usage: $0 <number_of_workers> <routing_strategy> [model_name] [endpoint_name]"
     echo "Error: Must specify at least number of workers and routing strategy"
     echo "Optional: model_name (default: deepseek-ai/DeepSeek-R1-Distill-Llama-8B)"
-    echo "Optional: endpoint_name (default: dynemo.process.chat/completions)"
+    echo "Optional: endpoint_name (default: dynamo.process.chat/completions)"
     exit 1
 fi
 
 NUM_WORKERS=$1
 ROUTING_STRATEGY=$2
 MODEL_NAME=${3:-"deepseek-ai/DeepSeek-R1-Distill-Llama-8B"}
-ENDPOINT_NAME=${4:-"dynemo.process.chat/completions"}
+ENDPOINT_NAME=${4:-"dynamo.process.chat/completions"}
 VALID_STRATEGIES=("prefix")
 SESSION_NAME="v"
 WORKDIR="/workspace/examples/python_rs/llm/vllm"

examples/python_rs/llm/vllm/sdk_basic_service/README.md

@@ -26,12 +26,12 @@ Users/Clients (HTTP)
 ┌─────────────┐
 │  Frontend   │  HTTP API endpoint (/generate)
 └─────────────┘
-      │ dynemo/distributed-runtime
+      │ dynamo/runtime
       ▼
 ┌─────────────┐
 │   Middle    │
 └─────────────┘
-      │ dynemo/distributed-runtime
+      │ dynamo/runtime
       ▼
 ┌─────────────┐
 │  Backend    │
@@ -45,7 +45,7 @@ Users/Clients (HTTP)
 ```bash
 cd /workspace/examples/python_rs/llm/vllm
 
-dynemo-sdk serve sdk_basic_service.basic:Frontend
+dynamo-sdk serve sdk_basic_service.basic:Frontend
 ```
 
 2. Send request to frontend using curl -
@@ -58,4 +58,4 @@ curl -X 'POST' \
   -d '{
   "text": "test"
 }'
-```
+```
\ No newline at end of file

examples/python_rs/llm/vllm/sdk_basic_service/basic.py

@@ -15,7 +15,7 @@
 
 from pydantic import BaseModel
 
-from dynemo.sdk import api, depends, dynemo_endpoint, service
+from dynamo.sdk import api, depends, dynamo_endpoint, service
 
 """
 Pipeline Architecture:
@@ -26,12 +26,12 @@ Users/Clients (HTTP)
 ┌─────────────┐
 │  Frontend   │  HTTP API endpoint (/generate)
 └─────────────┘
-      │ dynemo/distributed-runtime
+      │ dynamo/runtime
       ▼
 ┌─────────────┐
 │   Middle    │
 └─────────────┘
-      │ dynemo/distributed-runtime
+      │ dynamo/runtime
       ▼
 ┌─────────────┐
 │  Backend    │
@@ -50,7 +50,7 @@ class ResponseType(BaseModel):
 @service(
     resources={"cpu": "2"},
     traffic={"timeout": 30},
-    dynemo={
+    dynamo={
         "enabled": True,
         "namespace": "inference",
     },
@@ -60,9 +60,10 @@ class Backend:
     def __init__(self) -> None:
         print("Starting backend")
 
-    @dynemo_endpoint()
+    @dynamo_endpoint()
     async def generate(self, req: RequestType):
         """Generate tokens."""
+        print("here2")
         req_text = req.text
         print(f"Backend received: {req_text}")
         text = f"{req_text}-back"
@@ -73,7 +74,7 @@ class Backend:
 @service(
     resources={"cpu": "2"},
     traffic={"timeout": 30},
-    dynemo={"enabled": True, "namespace": "inference"},
+    dynamo={"enabled": True, "namespace": "inference"},
 )
 class Middle:
     backend = depends(Backend)
@@ -81,13 +82,14 @@ class Middle:
     def __init__(self) -> None:
         print("Starting middle")
 
-    @dynemo_endpoint()
+    @dynamo_endpoint()
     async def generate(self, req: RequestType):
         """Forward requests to backend."""
         req_text = req.text
         print(f"Middle received: {req_text}")
         text = f"{req_text}-mid"
         next_request = RequestType(text=text).model_dump_json()
+        print("here5")
         async for response in self.backend.generate(next_request):
             print(f"Middle received response: {response}")
             yield f"Middle: {response}"

examples/python_rs/llm/vllm_nixl/README.md

@@ -33,7 +33,7 @@ All of the commands below are run inside the same container.
 
 ## Run deployment
 
-Add model to dynemo and start http server.
+Add model to dynamo and start http server.
 
 In terminal 0:
 ```
@@ -65,7 +65,7 @@ CUDA_VISIBLE_DEVICES=0 python prefill_worker.py \
     --model deepseek-ai/DeepSeek-R1-Distill-Llama-8B \
     --enforce-eager \
     --kv-transfer-config \
-    '{"kv_connector":"DynemoNixlConnector"}'
+    '{"kv_connector":"DynamoNixlConnector"}'
 ```
 
 In terminal 2:
@@ -78,7 +78,7 @@ CUDA_VISIBLE_DEVICES=1,2 python3 worker.py \
     --enforce-eager \
     --tensor-parallel-size 2 \
     --kv-transfer-config \
-    '{"kv_connector":"DynemoNixlConnector"}'
+    '{"kv_connector":"DynamoNixlConnector"}'
 ```
 
 
@@ -157,7 +157,7 @@ rm -r /tmp/nixl
 - [x] Zero copy
 - [x] Conditional remote prefill
 - [x] Manual example with tp > 1
-- [x] Run on dynemo distributed runtime
+- [x] Run on dynamo distributed runtime
 - [x] add oai http endpoint
 - [x] Sample only on decode, do note return remote prefill response
 - [x] Check if all transfers finished before moving to decode

examples/python_rs/llm/vllm_nixl/common.py

@@ -22,7 +22,7 @@ from vllm.distributed.device_communicators.nixl import NixlMetadata
 from vllm.engine.arg_utils import AsyncEngineArgs
 from vllm.utils import FlexibleArgumentParser
 
-from dynemo.runtime import DistributedRuntime
+from dynamo.runtime import DistributedRuntime
 
 METADATA_DIR = "/tmp/nixl"
 

examples/python_rs/llm/vllm_nixl/prefill_worker.py

@@ -26,7 +26,7 @@ from vllm.entrypoints.openai.api_server import (
 from vllm.inputs.data import TokensPrompt
 from vllm.remote_prefill import RemotePrefillParams, RemotePrefillRequest
 
-from dynemo.runtime import DistributedRuntime, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_worker
 
 
 class RequestHandler:
@@ -71,7 +71,7 @@ class RequestHandler:
             yield
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     component = runtime.namespace("test-nixl").component("prefill")
     await component.create_service()

examples/python_rs/llm/vllm_nixl/worker.py

@@ -33,7 +33,7 @@ from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
 from vllm.entrypoints.openai.serving_models import BaseModelPath, OpenAIServingModels
 from vllm.remote_prefill import RemotePrefillParams, RemotePrefillRequest
 
-from dynemo.runtime import DistributedRuntime, dynemo_endpoint, dynemo_worker
+from dynamo.runtime import DistributedRuntime, dynamo_endpoint, dynamo_worker
 
 
 class RequestHandler:
@@ -83,7 +83,7 @@ class RequestHandler:
 
         return callback
 
-    @dynemo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
+    @dynamo_endpoint(ChatCompletionRequest, ChatCompletionStreamResponse)
     async def generate(self, request):
         if not self.initialized:
             await self.init()
@@ -109,7 +109,7 @@ class RequestHandler:
             yield response
 
 
-@dynemo_worker()
+@dynamo_worker()
 async def worker(runtime: DistributedRuntime, engine_args: AsyncEngineArgs):
     component = runtime.namespace("test-nixl").component("vllm")
     await component.create_service()

examples/rust/Cargo.lock

@@ -956,7 +956,7 @@ dependencies = [
 ]
 
 [[package]]
-name = "dynemo-llm"
+name = "dynamo-llm"
 version = "0.2.1"
 dependencies = [
  "anyhow",
@@ -971,7 +971,7 @@ dependencies = [
  "chrono",
  "cmake",
  "derive_builder",
- "dynemo-runtime",
+ "dynamo-runtime",
  "either",
  "erased-serde",
  "futures",
@@ -1005,7 +1005,7 @@ dependencies = [
 ]
 
 [[package]]
-name = "dynemo-runtime"
+name = "dynamo-runtime"
 version = "0.2.1"
 dependencies = [
  "anyhow",
@@ -1463,7 +1463,7 @@ checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
 name = "hello_world"
 version = "0.2.0"
 dependencies = [
- "dynemo-runtime",
+ "dynamo-runtime",
 ]
 
 [[package]]
@@ -1488,8 +1488,8 @@ name = "http"
 version = "0.2.0"
 dependencies = [
  "clap",
- "dynemo-llm",
- "dynemo-runtime",
+ "dynamo-llm",
+ "dynamo-runtime",
  "serde",
  "serde_json",
  "tokio",
@@ -1988,8 +1988,8 @@ name = "llmctl"
 version = "0.2.0"
 dependencies = [
  "clap",
- "dynemo-llm",
- "dynemo-runtime",
+ "dynamo-llm",
+ "dynamo-runtime",
  "serde",
  "serde_json",
  "tabled",
@@ -3390,7 +3390,7 @@ dependencies = [
 name = "service_metrics"
 version = "0.2.0"
 dependencies = [
- "dynemo-runtime",
+ "dynamo-runtime",
  "futures",
  "serde",
  "serde_json",