Feature/video gen#20
Merged
cryptopoly merged 4 commits intomainfrom May 1, 2026
Merged
Conversation
The warm-pool eviction policy was a fixed count (MAX_WARM_MODELS = 2), so loading three 17-18 GB models in succession on a 64 GB Mac kept all three resident — 17 + 17 + 18 + 21 GB llama-server = ~73 GB, swap, OS crash. Replace the count cap with a memory-aware policy: - _model_resident_bytes(info): on-disk weight size as a proxy for RAM, works for both mlx-lm (mmap) and llama.cpp (full-load). - _memory_budget_bytes(): live psutil snapshot of available RAM minus WARM_POOL_MEMORY_HEADROOM_BYTES (6 GB, mirrors spareHeadroomGb). - _evict_warm_pool(incoming_bytes=N): apply the count cap first (defensive — handles psutil-unavailable hosts), then evict oldest until pool + incoming fits under the budget. - _park_active_engine_or_unload now passes the parked model's resident bytes so eviction targets the right footprint. Chat picker also defaulted to a catalog-only model (nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF) on a fresh thread, then 500'd on Load with 'isn't downloaded on this machine'. Two fixes: - src/App.tsx threadModelOptions drops the catalog branch entirely. Discover tab is the place to pull a new model; the picker shouldn't surface entries that don't exist locally. - backend_service/state.py _default_session_model now prefers the first available text library entry over _default_chat_variant() when no model is loaded. Falls back to the catalog default only if the library is empty (true first-launch case).
- Convert image/video model catalog views to denser sortable rows with filters - Replace model actions/status controls with icon buttons and tooltips - Add shared video component delete handling and filter non-chat models out of chat selectors - Add collapsed install logs for launch-modal custom runtime strategies - Calibrate image/video RAM estimates with runtime and device-specific footprint metadata - Cover catalog payloads, discover sorting, library filtering, and safety estimators with tests
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.