docker-compose.yml

# Project-local Ollama for the retail feedback demo.
#
# FULLY ISOLATED from the mikkonumminendev RAG stack: own model volume, models
# pulled into it once (~10 GB). Nothing this demo does may touch the live RAG's
# model store. The GPU is still shared with that stack:
#   - Run inference ONLY while the RAG stack is down (announce first — see CLAUDE.md).
#   - No restart policy on purpose: must never auto-start and contend for the GPU.
#
# One-time model pull (network + disk only, no GPU):
#   docker compose up -d ollama
#   docker compose exec ollama ollama pull hf.co/mradermacher/Llama-Poro-2-8B-Instruct-GGUF:Q4_K_M
#   docker compose exec ollama ollama pull qwen3:8b
#   docker compose stop ollama
services:
  ollama:
    image: ollama/ollama:latest
    container_name: retail-rag-ollama
    environment:
      # Served context window — a SERVER-side knob (per-request num_ctx is not
      # set anywhere). 4096 fits Phase 0 structuring prompts; revisit for the
      # Phase 4 synthesis window. Larger costs VRAM per loaded model.
      OLLAMA_CONTEXT_LENGTH: ${OLLAMA_CONTEXT_LENGTH:-4096}
      # Concurrency MUST agree with the app-side LlmGate (Ingest:LlmMaxConcurrency
      # = 2, shed-not-queue), or the two layers fight: the gate fast-sheds at 2
      # while Ollama silently deep-queues under it (default OLLAMA_MAX_QUEUE=512),
      # so a call the gate admits can still stall for minutes in a hidden backlog
      # bounded only by Ingest:LlmCallTimeoutMs. NUM_PARALLEL=2 lets both admitted
      # slots actually run in parallel instead of being serialized; a SHALLOW
      # queue makes Ollama shed too rather than hide a deep one (ADR-0040). Safe
      # on VRAM because this container runs only while the shared RAG is down, so
      # the whole GPU is ours. Tunable via env if a future GPU is tighter.
      OLLAMA_NUM_PARALLEL: ${OLLAMA_NUM_PARALLEL:-2}
      OLLAMA_MAX_QUEUE: ${OLLAMA_MAX_QUEUE:-8}
    ports:
      - "11434:11434"
    healthcheck:
      test: ["CMD", "ollama", "list"]
      interval: 10s
      timeout: 5s
      retries: 12
    volumes:
      - ollama-models:/root/.ollama
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

volumes:
  ollama-models:
    name: retail-rag-ollama-models