Build Order¶

This appendix details the recommended component build order derived from architecture research. The order is determined by dependency analysis — each component is built only after its dependencies exist. The implementation plan phases reference this appendix for the detailed dependency chain.

Step	Component	Status	Depends On	What It Unblocks
1	libs/adapter-registry	✅ Implemented	Nothing	All components that store or retrieve adapters
2	services/lora-server	🔄 Replaced	adapter-registry	Replaced by libs/inference providers (TransformersProvider, LlamaCppProvider, OllamaProvider, VLLMProvider)
3	libs/model-training (extend)	✅ Implemented	inference providers	Hypernetwork, D2L pipeline, TIES/DARE merging, QLoRA trainer
4	services/api-service (extend)	✅ Implemented	adapter-registry	REST API for adapter management
5	services/rune-agent	✅ Implemented	inference, api-service, sandbox	LangGraph state graph; 4-phase pipeline in scripts/rune_runner.py
6	services/evolution-svc	⚠️ Partial	adapter-registry, inference	REST stubs; evolution logic in scripts/swarm_evolution.py
7	services/training-svc	✅ Implemented	adapter-registry, model-training	POST /train/lora, POST /train/hypernetwork, GET /jobs/{id}
8	Hypernetwork	✅ Implemented	training-svc, adapter corpus	DocToLoraHypernetwork + D2L training pipeline
—	scripts/ orchestration	✅ Implemented	All above	rune_runner.py, swarm.py, swarm_workers.py, swarm_evolution.py

Dependency Graph¶

flowchart TD
    adapter_registry["adapter-registry ✅"] --> inference_providers["inference providers ✅"]
    adapter_registry --> api_service["api-service ✅"]
    inference_providers --> model_training["model-training ✅"]
    inference_providers --> rune_agent["rune-agent ✅"]
    api_service --> rune_agent
    adapter_registry --> evolution_svc["evolution-svc ⚠️"]
    inference_providers --> evolution_svc
    adapter_registry --> training_svc["training-svc ✅"]
    model_training --> training_svc
    training_svc --> hypernetwork["hypernetwork ✅"]
    rune_agent --> hypernetwork
    rune_agent --> scripts["scripts/ orchestration ✅"]
    training_svc --> scripts
    evolution_svc --> scripts