fix inference stdout and README docker port

- inference.py: replace stray print() with emit_log(END) so all stdout
lines are structured [START]/[STEP]/[END] — fixes failing inference unit test
- inference.py: fix docstring ENV_URL default 8000 -> 7860
- README.md: fix docker run example -p 8000:7860 -> -p 7860:7860 so
default ENV_URL (localhost:7860) hits the container without extra config

README.md

@@ -349,13 +349,19 @@ Build:
 docker build -f server/Dockerfile -t helpdesk-ticket-routing .
 ```
 
-Run locally while keeping the inference default URL unchanged:
+Run locally:
 
 ```bash
-docker run -p 8000:7860 helpdesk-ticket-routing
+docker run -p 7860:7860 helpdesk-ticket-routing
 ```
 
-If you instead publish the container on another port, set `ENV_URL` accordingly before running `inference.py`.
+Then run inference against it (default `ENV_URL` points to `http://localhost:7860`):
+
+```bash
+python inference.py
+```
+
+If you publish the container on a different host port, set `ENV_URL` accordingly before running `inference.py`.
 
 If local Docker is blocked by machine setup, the repo also includes a GitHub Actions smoke test at `.github/workflows/docker-smoke-test.yml`. That workflow builds the image on a GitHub-hosted runner, starts the container, checks `/health` and `/tasks`, and runs heuristic `inference.py` against the container.
 

inference.py

@@ -6,7 +6,7 @@ Environment variables
 ---------------------
 ENV_URL
     Base URL of the running OpenEnv server.
-    Default: ``http://localhost:8000``
+    Default: ``http://localhost:7860``
 
 API_BASE_URL
     LLM provider base URL (OpenAI-compatible endpoint).
@@ -404,7 +404,7 @@ def run() -> None:
         if task_id in all_results
     ]
     overall_avg = round(sum(overall) / len(overall), 4) if overall else 0.0
-    print(f"Overall average reward: {overall_avg:.4f}")
+    emit_log("END", overall_avg=overall_avg, tasks_completed=len(overall))
 
 
 if __name__ == "__main__":