Merge pull request #29 from zhaochenyang20/upd_docs

[Docs] Highlight the launch with yaml

Author: zhaochenyang20

.gitignore

@@ -3,6 +3,8 @@ __pycache__/
 *.py[codz]
 *$py.class
 
+output.png
+
 # C extensions
 *.so
 

README.md

@@ -41,6 +41,26 @@ cd ..
 
 ## Quick Start
 
+### Co-launch workers and router via YAML config
+
+Instead of starting workers manually, you can let the router spawn and manage them via a YAML config file.
+
+```bash
+sglang-d-router --port 30081 --launcher-config examples/local_launcher.yaml
+```
+
+```yaml
+launcher:
+  backend: local
+  model: Qwen/Qwen-Image
+  num_workers: 8
+  num_gpus_per_worker: 1
+  worker_base_port: 10090
+  wait_timeout: 600
+```
+
+### Manual Launch Workers and Connect to Router
+
 ```bash
 # If connect to HuggingFace is not allowed
 # You can set the environment variable SGLANG_USE_MODELSCOPE=TRUE
@@ -65,7 +85,6 @@ sglang-d-router --port 30081 \
 
 ## Demonstrative Examples
 
-
 ### With Python Requests
 
 ```python
@@ -98,12 +117,6 @@ with open("output.png", "wb") as f:
     f.write(img)
 print("Saved to output.png")
 
-# Video generation request
-resp = requests.post(f"{ROUTER}/generate_video", json={
-    "model": "Qwen/Qwen-Image",
-    "prompt": "a flowing river",
-})
-print(resp.json())
 
 # Check per-worker health and load
 resp = requests.get(f"{ROUTER}/health_workers")
@@ -158,30 +171,6 @@ curl -X POST http://localhost:30081/update_weights_from_disk \
     -d '{"model_path": "Qwen/Qwen-Image-2512"}'
 ```
 
-### Auto-launch workers via YAML config
-
-Instead of starting workers manually, you can let the router spawn and manage
-them through a launcher backend.
-
-**Local subprocess launcher** (`examples/local_launcher.yaml`):
-
-```bash
-sglang-d-router --port 30081 --launcher-config examples/local_launcher.yaml
-```
-
-```yaml
-launcher:
-  backend: local
-  model: Qwen/Qwen-Image
-  num_workers: 2
-  num_gpus_per_worker: 1
-  worker_base_port: 10090
-  wait_timeout: 600
-```
-
-Fields not set in the YAML fall back to defaults defined in each backend's
-config dataclass (see `LocalLauncherConfig`).
-
 ## Acknowledgment
 
 This project is derived from [radixark/miles#544](https://github.com/radixark/miles/pull/544). Thanks to the original authors.

development.md

@@ -23,9 +23,9 @@ Single benchmark:
 ```bash
 python tests/benchmarks/diffusion_router/bench_router.py \
     --model Qwen/Qwen-Image \
-    --num-workers 2 \
+    --num-workers 8 \
     --num-prompts 20 \
-    --max-concurrency 4
+    --max-concurrency 8
 ```
 
 Algorithm comparison:

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sglang-diffusion-routing"
-version = "0.1.0"
+version = "0.1.1"
 description = "Load-balancing router for SGLang diffusion workers"
 readme = "README.md"
 requires-python = ">=3.10"