AI Models in RawCull

Sun, 20 Sep 2026 00:00:00 +0000

AI Models in RawCull

RawCull uses several local machine-learning backends, but it does not treat them as interchangeable. Each model family has a deliberately narrow job:

Model or backend	RawCull job	Output used by RawCull
DataComp CLIP	Image similarity, burst grouping, semantic search, and coarse subject labels for Deep Review	Normalized image/text embedding vectors and cosine distances/similarities
OpenAI CLIP	Fully implemented alternative CLIP bundle; currently excluded from the production model list	The same typed CLIP artifacts as DataComp, with a different model fingerprint
SAM 3	Prompted subject segmentation for Deep Review and separate instance segmentation for Objects	A chosen subject mask, or up to eight numbered masks per concept
Qwen3-VL-2B-Instruct	Standalone photo assessment; concept discovery and board interpretation in Objects	A photo assessment, validated object concepts and per-object findings, or a visible retryable response failure
Apple Vision feature print	Always-available image-similarity fallback	Opaque Vision feature-print artifacts and native distances

All inference stays in the application process. The downloadable model assets are installed separately because they are large, but the analysis path does not send photographs to a remote inference service.

The RawCull AI Runtime

Thu, 03 Sep 2026 00:00:00 +0000

The RawCull AI Runtime

RawCull’s AI runtime is the long-lived object graph that connects downloaded model assets to stable application features. It is not one model, one thread, or a background daemon. It is a set of objects with deliberately different lifetimes and actor-isolation rules.

The current implementation has two runtime layers:

Runtime	Primary responsibility
`RawCullAIModelRuntime`	Own concrete provider/resource lifecycles: CLIP, SAM 3, Qwen, Vision, model capability snapshots, separate subject/object mask stores, and segmentation-service installation.
`RawCullIntelligenceRuntime`	Own stable similarity, semantic search, Deep Review, Qwen, and Objects feature lifetimes; apply complete, revisioned similarity/semantic/segmentation settings decisions without rebuilding the graph.

That distinction replaces the older, broader RawCullAIIntegration shape. The authoritative sources are RawCullAIModelRuntime.swift and RawCullIntelligenceRuntime.swift. For model algorithms and data products, see AI Models in RawCull.

Objects on RawCull

AI Models in RawCull

AI Models in RawCull

The RawCull AI Runtime

The RawCull AI Runtime