Add Render Blueprint for hosted demo

README.md

@@ -60,6 +60,8 @@ python scripts/smoke_check.py
 
 Deployment notes are in `docs/deployment.md`.
 
+There is also a Render Blueprint in `render.yaml` for a hosted demo with one public gateway, two private workers, and a managed cache.
+
 ## Benchmarks
 
 The scenario runner starts local gateway and worker processes, sends load, then writes JSON and markdown results.

docs/deployment.md

@@ -42,3 +42,15 @@ CACHE_ENABLED=true
 This is not a production ML platform yet. It is a compact inference mesh that demonstrates training, serving, caching, routing, live worker controls, failure recovery, and benchmark evidence.
 
 The next real hosting step is choosing a provider and wiring its private service networking. The Docker setup here is intentionally provider-neutral so the app can go to Fly.io, Render, Railway, a VM, or a Kubernetes cluster without changing the Python code.
+
+## Render Blueprint
+
+`render.yaml` defines a Render Blueprint for:
+
+- one public gateway web service
+- two private worker services
+- one Render Key Value cache
+
+The gateway reads `WORKER_A_HOSTPORT` and `WORKER_B_HOSTPORT`, which Render fills from each private worker service. That avoids hardcoding internal hostnames.
+
+Important: the private worker services use Render's `starter` plan in the Blueprint because private services do not run on the free plan. Review the cost in Render before applying the Blueprint.

gateway/Dockerfile

@@ -9,4 +9,4 @@ RUN pip install --no-cache-dir -r requirements.txt
 
 COPY gateway ./gateway
 HEALTHCHECK --interval=30s --timeout=3s --retries=3 CMD python -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8000/health', timeout=2).read()"
-CMD ["uvicorn", "gateway.app.main:app", "--host", "0.0.0.0", "--port", "8000"]
+CMD ["sh", "-c", "uvicorn gateway.app.main:app --host 0.0.0.0 --port ${PORT:-8000}"]

gateway/app/main.py

@@ -23,9 +23,20 @@ def env_flag(name: str, default: bool = True) -> bool:
     return value.lower() not in {"0", "false", "no", "off"}
 
 
+def parse_worker_hostports(env: dict[str, str] | None = None) -> list[str]:
+    values = env or os.environ
+    urls = []
+    for key in sorted(values):
+        if key.startswith("WORKER_") and key.endswith("_HOSTPORT"):
+            hostport = values[key].strip()
+            if hostport:
+                urls.append(f"http://{hostport}")
+    return urls
+
+
 def parse_worker_urls(raw: str | None) -> list[str]:
     if not raw:
-        return ["http://localhost:8011"]
+        return parse_worker_hostports() or ["http://localhost:8011"]
     return [item.strip() for item in raw.split(",") if item.strip()]
 
 

render.yaml

@@ -0,0 +1,55 @@
+services:
+  - type: keyvalue
+    name: modelmesh-cache
+    plan: free
+    ipAllowList: []
+
+  - type: pserv
+    runtime: docker
+    name: modelmesh-worker-a
+    plan: starter
+    dockerfilePath: ./worker/Dockerfile
+    dockerContext: .
+    envVars:
+      - key: WORKER_ID
+        value: worker-a
+
+  - type: pserv
+    runtime: docker
+    name: modelmesh-worker-b
+    plan: starter
+    dockerfilePath: ./worker/Dockerfile
+    dockerContext: .
+    envVars:
+      - key: WORKER_ID
+        value: worker-b
+
+  - type: web
+    runtime: docker
+    name: modelmesh-gateway
+    plan: starter
+    dockerfilePath: ./gateway/Dockerfile
+    dockerContext: .
+    healthCheckPath: /health
+    envVars:
+      - key: WORKER_A_HOSTPORT
+        fromService:
+          type: pserv
+          name: modelmesh-worker-a
+          property: hostport
+      - key: WORKER_B_HOSTPORT
+        fromService:
+          type: pserv
+          name: modelmesh-worker-b
+          property: hostport
+      - key: REDIS_URL
+        fromService:
+          type: keyvalue
+          name: modelmesh-cache
+          property: connectionString
+      - key: REQUEST_TIMEOUT_SECONDS
+        value: "1.5"
+      - key: ROUTER_STRATEGY
+        value: load_aware
+      - key: CACHE_ENABLED
+        value: "true"

tests/test_cache.py

@@ -1,4 +1,5 @@
 from gateway.app.cache import PredictionCache
+from gateway.app.main import parse_worker_hostports, parse_worker_urls
 
 
 def test_memory_cache_round_trip() -> None:
@@ -27,3 +28,17 @@ def test_disabled_cache_is_a_noop() -> None:
     cache.set(key, {"prediction": {"label": "healthy"}})
 
     assert cache.get(key) is None
+
+
+def test_worker_hostport_env_vars_are_render_friendly() -> None:
+    env = {
+        "WORKER_A_HOSTPORT": "worker-a:9000",
+        "WORKER_B_HOSTPORT": "worker-b:9000",
+        "OTHER_VALUE": "ignored",
+    }
+
+    assert parse_worker_hostports(env) == ["http://worker-a:9000", "http://worker-b:9000"]
+
+
+def test_worker_urls_still_accepts_explicit_list() -> None:
+    assert parse_worker_urls("http://a:9000,http://b:9000") == ["http://a:9000", "http://b:9000"]

worker/Dockerfile

@@ -9,4 +9,4 @@ RUN pip install --no-cache-dir -r requirements.txt
 
 COPY worker ./worker
 HEALTHCHECK --interval=30s --timeout=3s --retries=3 CMD python -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:9000/health', timeout=2).read()"
-CMD ["uvicorn", "worker.app.main:app", "--host", "0.0.0.0", "--port", "9000"]
+CMD ["sh", "-c", "uvicorn worker.app.main:app --host 0.0.0.0 --port ${PORT:-9000}"]