feat(01-02): LLM Backend Pool — LiteLLM Router with Ollama + Anthropic + OpenAI fallback

- Create llm_pool/router.py: LiteLLM Router with fast (Ollama) and quality (Anthropic/OpenAI) model groups
- Configure fallback chain: quality providers fail -> fast group
- Pin LiteLLM to ==1.82.5 (avoid September 2025 OOM regression in later releases)
- Create llm_pool/main.py: FastAPI service on port 8004 with /complete and /health endpoints
- Add providers/__init__.py: reserved for future per-provider customization
- Update docker-compose.yml: add llm-pool and celery-worker service stubs

docker-compose.yml

@@ -64,3 +64,72 @@ services:
               capabilities: [gpu]
     # Service starts even if no GPU is available — GPU config is optional
     restart: unless-stopped
+
+  llm-pool:
+    build:
+      context: .
+      dockerfile_inline: |
+        FROM python:3.12-slim
+        WORKDIR /app
+        RUN pip install uv
+        COPY pyproject.toml ./
+        COPY packages/shared ./packages/shared
+        COPY packages/llm-pool ./packages/llm-pool
+        RUN uv pip install --system -e packages/shared -e packages/llm-pool
+        CMD ["uvicorn", "llm_pool.main:app", "--host", "0.0.0.0", "--port", "8004"]
+    container_name: konstruct-llm-pool
+    ports:
+      - "8004:8004"
+    networks:
+      - konstruct-net
+    depends_on:
+      ollama:
+        condition: service_started
+      redis:
+        condition: service_healthy
+    environment:
+      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY:-}
+      - OPENAI_API_KEY=${OPENAI_API_KEY:-}
+      - OLLAMA_BASE_URL=http://ollama:11434
+      - REDIS_URL=redis://redis:6379/0
+      - LOG_LEVEL=INFO
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD-SHELL", "curl -sf http://localhost:8004/health || exit 1"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
+  celery-worker:
+    build:
+      context: .
+      dockerfile_inline: |
+        FROM python:3.12-slim
+        WORKDIR /app
+        RUN pip install uv
+        COPY pyproject.toml ./
+        COPY packages/shared ./packages/shared
+        COPY packages/orchestrator ./packages/orchestrator
+        RUN uv pip install --system -e packages/shared -e packages/orchestrator
+        CMD ["celery", "-A", "orchestrator.main", "worker", "--loglevel=info"]
+    container_name: konstruct-celery-worker
+    networks:
+      - konstruct-net
+    depends_on:
+      redis:
+        condition: service_healthy
+      postgres:
+        condition: service_healthy
+      llm-pool:
+        condition: service_healthy
+    environment:
+      - DATABASE_URL=postgresql+asyncpg://konstruct_app:konstruct_dev@postgres:5432/konstruct
+      - REDIS_URL=redis://redis:6379/0
+      - CELERY_BROKER_URL=redis://redis:6379/1
+      - CELERY_RESULT_BACKEND=redis://redis:6379/2
+      - LLM_POOL_URL=http://llm-pool:8004
+      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY:-}
+      - OPENAI_API_KEY=${OPENAI_API_KEY:-}
+      - OLLAMA_BASE_URL=http://ollama:11434
+      - LOG_LEVEL=INFO
+    restart: unless-stopped