refactor: rename CF_ORCH_URL → GPU_SERVER_URL (backward-compat alias kept)

GPU_SERVER_URL is the self-explanatory name a self-hoster can understand
without knowing CircuitForge internals. CF_ORCH_URL continues to work as
a drop-in fallback alias (runner.py, main.py both check GPU_SERVER_URL
first, then CF_ORCH_URL).

Updated everywhere the env var is referenced or documented:
- app/tasks/runner.py
- api/main.py
- app/llm/router.py
- .env.example (alias note added)
- compose.override.yml
- compose.cloud.yml
- config/llm.cloud.yaml
- tests/test_tasks/test_runner.py (primary key updated; 13/13 still pass)

Follows the GPU_SERVER_URL convention established in kiwi (see kiwi
app/core/config.py).

Closes: #55

.env.example

@@ -98,14 +98,16 @@ CF_APP_NAME=snipe
 # OLLAMA_HOST=http://localhost:11434
 # OLLAMA_MODEL=llava:7b
 
-# CF Orchestrator — routes vision/LLM tasks to a cf-orch coordinator for VRAM management.
+# GPU Server — routes vision/LLM tasks to a cf-orch coordinator for VRAM management.
 # Self-hosted: point at a local cf-orch coordinator if you have one running.
 # Cloud (internal): managed coordinator at orch.circuitforge.tech.
 # Leave unset to run vision tasks inline (no VRAM coordination).
-# CF_ORCH_URL=http://10.1.10.71:7700
+# GPU_SERVER_URL=http://10.1.10.71:7700
+#
+# CF_ORCH_URL is accepted as a backward-compat alias for GPU_SERVER_URL.
 #
 # cf-orch agent (compose --profile orch) — coordinator URL for the sidecar agent.
-# Defaults to CF_ORCH_URL if unset.
+# Defaults to GPU_SERVER_URL if unset.
 # CF_ORCH_COORDINATOR_URL=http://10.1.10.71:7700
 
 # ── Community DB (optional) ──────────────────────────────────────────────────

api/main.py

@@ -209,7 +209,7 @@ async def _lifespan(app: FastAPI):
             _category_cache.refresh(token_manager=None)  # bootstrap fallback
 
         try:
-            cforch_url = os.getenv("CF_ORCH_URL") or None
+            cforch_url = os.getenv("GPU_SERVER_URL") or os.getenv("CF_ORCH_URL") or None
             if cforch_url:
                 _query_translator = QueryTranslator(
                     category_cache=_category_cache,

app/llm/router.py

@@ -6,7 +6,7 @@ Snipe LLMRouter shim — tri-level config path priority.
 Config lookup order:
   1. <repo>/config/llm.yaml  — per-install local override
   2. ~/.config/circuitforge/llm.yaml  — user-level config (circuitforge-core default)
-  3. env-var auto-config  (ANTHROPIC_API_KEY, OPENAI_API_KEY, OLLAMA_HOST, CF_ORCH_URL)
+  3. env-var auto-config  (ANTHROPIC_API_KEY, OPENAI_API_KEY, OLLAMA_HOST, GPU_SERVER_URL)
 """
 from pathlib import Path
 

app/tasks/runner.py

@@ -8,9 +8,9 @@ Current task types:
                            result to trust_scores.photo_analysis_json (Paid tier).
 
 Image assessment routing:
-    Cloud (CF_ORCH_URL set): allocates via cf-orch task endpoint
+    Cloud (GPU_SERVER_URL set): allocates via cf-orch task endpoint
         product=snipe, task=image_assessment.
-    Local (no CF_ORCH_URL) or TaskNotFound fallback: uses LLMRouter
+    Local (no GPU_SERVER_URL) or TaskNotFound fallback: uses LLMRouter
         with a vision-capable local backend (moondream2, llava, etc.).
 """
 from __future__ import annotations
@@ -135,7 +135,7 @@ def _run_trust_photo_analysis(
     if listing_title:
         user_prompt = f"Assess this eBay listing image: {listing_title}"
 
-    cforch_url = os.getenv("CF_ORCH_URL")
+    cforch_url = os.getenv("GPU_SERVER_URL") or os.getenv("CF_ORCH_URL")
     if cforch_url:
         raw = _assess_via_orch(cforch_url, image_data_url, user_prompt)
     else:

compose.cloud.yml

@@ -20,9 +20,9 @@ services:
       CLOUD_MODE: "true"
       CLOUD_DATA_ROOT: /devl/snipe-cloud-data
       # DIRECTUS_JWT_SECRET, HEIMDALL_URL, HEIMDALL_ADMIN_TOKEN — set in .env (never commit)
-      # CF_ORCH_URL routes LLM query builder through cf-orch for VRAM-aware scheduling.
+      # GPU_SERVER_URL routes LLM query builder through cf-orch for VRAM-aware scheduling.
       # Override in .env to use a different coordinator URL.
-      CF_ORCH_URL: "http://host.docker.internal:7700"
+      GPU_SERVER_URL: "http://host.docker.internal:7700"
       CF_APP_NAME: snipe
     extra_hosts:
       - "host.docker.internal:host-gateway"

compose.override.yml

@@ -18,8 +18,8 @@ services:
     environment:
       - RELOAD=true
       # Point the LLM/vision task scheduler at the local cf-orch coordinator.
-      # Only has effect when CF_ORCH_URL is set (uncomment in .env, or set inline).
-      # - CF_ORCH_URL=http://10.1.10.71:7700
+      # Only has effect when GPU_SERVER_URL is set (uncomment in .env, or set inline).
+      # - GPU_SERVER_URL=http://10.1.10.71:7700
 
   # cf-orch agent — routes trust_photo_analysis vision tasks to the GPU coordinator.
   # Only starts when you pass --profile orch:

config/llm.cloud.yaml

@@ -6,7 +6,7 @@
 # (claude_code, copilot) are intentionally excluded here.
 #
 # CF Orchestrator routes both ollama and vllm allocations for VRAM-aware
-# scheduling.  CF_ORCH_URL must be set in .env for allocations to resolve;
+# scheduling.  GPU_SERVER_URL must be set in .env for allocations to resolve;
 # if cf-orch is unreachable the backend falls back to its static base_url.
 #
 # Model choice for query builder: llama3.1:8b

tests/test_tasks/test_runner.py

@@ -173,7 +173,7 @@ def _make_orch_client_mock(vision_json: str) -> MagicMock:
 
 
 def test_run_task_photo_analysis_orch_success(tmp_db: Path):
-    """Cloud path: CFOrchClient.task_allocate is used when CF_ORCH_URL is set."""
+    """Cloud path: CFOrchClient.task_allocate is used when GPU_SERVER_URL is set."""
     task_id, _ = insert_task(tmp_db, "trust_photo_analysis", job_id=1, params=_PARAMS)
 
     chat_resp = MagicMock()
@@ -181,7 +181,7 @@ def test_run_task_photo_analysis_orch_success(tmp_db: Path):
     chat_resp.raise_for_status = MagicMock()
 
     with patch("app.tasks.runner.requests") as mock_req, \
-         patch.dict("os.environ", {"CF_ORCH_URL": "http://cf-orch.local:8700"}), \
+         patch.dict("os.environ", {"GPU_SERVER_URL": "http://cf-orch.local:8700"}), \
          patch("app.tasks.runner.httpx") as mock_httpx, \
          patch("circuitforge_orch.client.CFOrchClient") as MockClient:
 
@@ -216,7 +216,7 @@ def test_run_task_photo_analysis_orch_uses_image_assessment_task(tmp_db: Path):
     chat_resp.raise_for_status = MagicMock()
 
     with patch("app.tasks.runner.requests") as mock_req, \
-         patch.dict("os.environ", {"CF_ORCH_URL": "http://cf-orch.local:8700"}), \
+         patch.dict("os.environ", {"GPU_SERVER_URL": "http://cf-orch.local:8700"}), \
          patch("app.tasks.runner.httpx") as mock_httpx, \
          patch("circuitforge_orch.client.CFOrchClient") as MockClient:
 
@@ -248,7 +248,7 @@ def test_run_task_photo_analysis_orch_sends_image_url_content(tmp_db: Path):
         return resp
 
     with patch("app.tasks.runner.requests") as mock_req, \
-         patch.dict("os.environ", {"CF_ORCH_URL": "http://cf-orch.local:8700"}), \
+         patch.dict("os.environ", {"GPU_SERVER_URL": "http://cf-orch.local:8700"}), \
          patch("app.tasks.runner.httpx") as mock_httpx, \
          patch("circuitforge_orch.client.CFOrchClient") as MockClient:
 
@@ -282,7 +282,7 @@ def test_run_task_photo_analysis_orch_task_not_found_falls_back(tmp_db: Path):
     client_instance.task_allocate.return_value = cm
 
     with patch("app.tasks.runner.requests") as mock_req, \
-         patch.dict("os.environ", {"CF_ORCH_URL": "http://cf-orch.local:8700"}), \
+         patch.dict("os.environ", {"GPU_SERVER_URL": "http://cf-orch.local:8700"}), \
          patch("circuitforge_orch.client.CFOrchClient", return_value=client_instance), \
          patch("app.tasks.runner._assess_via_local_llm", return_value=_VISION_JSON) as mock_local: