Add LLM Judge container infrastructure

- Add cicd/docker-compose.judge.yml for stable reference Ollama
- Runs on port 11435 (separate from test subject on 11434)
- Uses dogkeeper886/ollama37:latest from DockerHub
- Add cicd/README.md documenting CI infrastructure

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

cicd/README.md

@@ -0,0 +1,54 @@
+# CI/CD Infrastructure
+
+This folder contains CI/CD infrastructure components separate from the main build system.
+
+## Components
+
+### LLM Judge (`docker-compose.judge.yml`)
+
+A stable reference Ollama instance for evaluating test results.
+
+**Purpose:**
+- Acts as secondary judge alongside simple exit-code checking
+- Analyzes test logs semantically to detect hidden issues
+- Uses stable DockerHub image (not the build being tested)
+
+**Architecture:**
+```
+Port 11434 → ollama37 (test subject, local build)
+Port 11435 → ollama37-judge (stable reference, DockerHub)
+```
+
+**Usage:**
+```bash
+# Start judge container
+cd cicd
+docker compose -f docker-compose.judge.yml up -d
+
+# Check status
+docker compose -f docker-compose.judge.yml ps
+
+# Pull model for judging (first time)
+curl -X POST http://localhost:11435/api/pull -d '{"name": "gemma3:1b"}'
+
+# Stop judge
+docker compose -f docker-compose.judge.yml down
+```
+
+## Folder Structure
+
+```
+cicd/
+├── docker-compose.judge.yml   # LLM Judge container
+├── README.md                  # This file
+└── scripts/                   # (future) CI helper scripts
+```
+
+## Related Components
+
+| Component | Location | Purpose |
+|-----------|----------|---------|
+| Test subject | `docker/docker-compose.yml` | Ollama build being tested |
+| Test runner | `tests/src/` | Executes tests, uses judge |
+| Test cases | `tests/testcases/` | YAML test definitions |
+| Workflows | `.github/workflows/` | CI pipeline definitions |

cicd/docker-compose.judge.yml

@@ -0,0 +1,33 @@
+services:
+  # LLM Judge - stable reference version for evaluating test results
+  # Runs on port 11435 to avoid conflict with test subject (11434)
+  ollama-judge:
+    image: dogkeeper886/ollama37:latest
+    container_name: ollama37-judge
+    runtime: nvidia
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
+    ports:
+      - "11435:11434"
+    volumes:
+      - ollama-judge-data:/root/.ollama
+    environment:
+      - OLLAMA_HOST=0.0.0.0:11434
+      - NVIDIA_VISIBLE_DEVICES=all
+      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD", "/usr/local/bin/ollama", "list"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 5s
+
+volumes:
+  ollama-judge-data:
+    name: ollama-judge-data