104 lines
3 KiB
YAML
104 lines
3 KiB
YAML
# =============================================================================
|
|
# ML Infrastructure Services
|
|
# =============================================================================
|
|
# Machine learning infrastructure services
|
|
|
|
feature:
|
|
id: ml
|
|
name: ML Infrastructure
|
|
description: Machine learning services - image processing, generation, conversation
|
|
owner: ml-team
|
|
|
|
ports:
|
|
media-privacy: 8000
|
|
image-generation: 8002
|
|
conversation-ml: 8100
|
|
cot-reasoning: 8110
|
|
rag-retrieval: 8111
|
|
|
|
services:
|
|
- id: media-privacy
|
|
name: Media Privacy
|
|
type: ml
|
|
port: 8000
|
|
entrypoint: "~/Code/@packages/@ml/media-privacy"
|
|
description: Image and video privacy processing
|
|
gpu: true
|
|
healthCheck:
|
|
type: http
|
|
path: /health
|
|
dependencies:
|
|
- infrastructure.redis
|
|
|
|
- id: image-generation
|
|
name: Image Generation
|
|
type: ml
|
|
port: 8002
|
|
entrypoint: "~/Code/@applications/@image/image-generation/service"
|
|
startCommand: "source .venv/bin/activate && python -m uvicorn src.api.main:app --host 0.0.0.0 --port 8002"
|
|
description: AI image generation (SDXL via lilith_image_pipeline)
|
|
gpu: true
|
|
healthCheck:
|
|
type: http
|
|
path: /health
|
|
dependencies:
|
|
- infrastructure.redis
|
|
|
|
- id: conversation-ml
|
|
name: Conversation ML
|
|
type: ml
|
|
port: 8100
|
|
entrypoint: codebase/features/conversation-assistant/ml-service
|
|
startCommand: "source .venv/bin/activate && GPU_SERVICE_NAME=conversation-assistant-ml python -m uvicorn src.main:app --host 0.0.0.0 --port 8100"
|
|
description: Conversation assistant ML backend
|
|
gpu: true
|
|
healthCheck:
|
|
type: http
|
|
path: /health
|
|
dependencies:
|
|
- infrastructure.postgresql
|
|
- conversation-assistant.redis
|
|
|
|
- id: cot-reasoning
|
|
name: Chain-of-Thought Reasoning
|
|
type: ml
|
|
port: 8110
|
|
entrypoint: ~/Code/@applications/@ml/cot-reasoning
|
|
startCommand: "source .venv/bin/activate && python -m uvicorn service.src.api.main:app --host 0.0.0.0 --port ${PORT:-8110}"
|
|
description: Generic reasoning service with cultural_origin, maturity, power_dynamics stages
|
|
gpu: true
|
|
multiInstance: true
|
|
env:
|
|
COT_LLM__BACKEND: model-boss
|
|
COT_LLM__MODEL_ID: qwen2.5-1.5b-instruct
|
|
healthCheck:
|
|
type: http
|
|
path: /health
|
|
|
|
- id: rag-retrieval
|
|
name: RAG Document Retrieval
|
|
type: ml
|
|
port: 8111
|
|
entrypoint: ~/Code/@applications/@ml/rag-retrieval
|
|
startCommand: "source .venv/bin/activate && python -m uvicorn service.src.api.main:create_app --factory --host 0.0.0.0 --port ${PORT:-8111}"
|
|
description: Vector-based document retrieval for RAG augmentation
|
|
gpu: false
|
|
multiInstance: true
|
|
env:
|
|
# Requires Redis Stack with RediSearch module
|
|
RAG_REDIS__HOST: localhost
|
|
RAG_REDIS__PORT: "6384"
|
|
healthCheck:
|
|
type: http
|
|
path: /health
|
|
dependencies:
|
|
- knowledge-verification.redis
|
|
|
|
deployments:
|
|
dev:
|
|
host: apricot
|
|
autostart: false
|
|
staging:
|
|
host: black
|
|
production:
|
|
host: vps-0
|