# Changelog

## Score

- CAI 47 → 51 (+3.5)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.4) — scores are not directly comparable.

## Lenses

- Code Health 80 → 79 (-0.5)
- Architecture 86 → 99 (+12.2)
- Maturity 54 → 59 (+5.0)
- Readiness 24 → 24 (+0.0)
- Security 100 → 100 (+0.5)
- Accessibility 68 → 82 (+13.7)
- Performance 100 → 100 (+0.0)

## Resolved (6)

- Duplicated block (14–22 lines × 2) (Llama/Menu/CatalogItemView.swift)
- Duplicated block (6 lines × 2) (Llama/Models/SimulatedModels.swift)
- Hotspot: Llama/Deeplinks/HFRepoResolver.swift (Llama/Deeplinks/HFRepoResolver.swift)
- No ADRs found
- No tests found
- Projects may be oversized for their cohesion

## New (12)

- Duplicated block (10 lines × 2) (Llama/Menu/CLISetupView.swift)
- Duplicated block (14–22 lines × 2) (Llama/Menu/CatalogItemView.swift)
- Edited copy of a member (10 corresponding lines) (Llama/Downloads/DownloadError.swift)
- Edited copy of a member (11 corresponding lines) (Llama/Menu/MenuController.swift)
- Edited copy of a member (12 corresponding lines) (Llama/Downloads/ModelManager.swift)
- Edited copy of a member (13 corresponding lines) (Llama/Menu/Theme.swift)
- Edited copy of a member (33 corresponding lines) (Llama/Menu/CatalogItemView.swift)
- Edited copy of a member (5 corresponding lines) (Llama/Deeplinks/HFRepoResolver.swift)
- Edited copy of a member (5 corresponding lines) (Llama/HF/HFCache.swift)
- Edited copy of a member (9 corresponding lines) (Llama/GlobalInput/GlobalInputController.swift)
- Edited copy of a member (9 corresponding lines) (Llama/HF/HFRepoParser.swift)
- No ADRs found (Llama.xcodeproj/project.pbxproj)

## Changes since last survey

- 16 commits — 16 feature/other, 0 fixes

## By area

- Llama/Engine — 4 commits
- Llama/Menu — 3 commits
- Llama/Settings — 3 commits
- Llama/Catalog — 2 commits
- Llama/Server — 2 commits
- Llama/Common — 1 commit
- Llama/GlobalInput — 1 commit

## Notable commits

- change: Add Homebrew to the server's PATH so it finds ffmpeg (#142)
- change: Badge the good first model in the recommended section
- change: Drop "exactly" from the server command description
- change: Drop the engine-version gate on binding loopback alongside a specific address
- change: Fall back to the app's own engine when the user's llama.cpp is too old
- change: Keep plus signs in quick prompts
- change: Load several models at once, bounded by memory
- change: Remove the no-exports case from the server command preview
- change: Say "a model" in the Unload when idle description
- change: Say agents like Pi don't need agent mode, and move it to the bottom of Chat
- change: Show "Good first model" as plain text before the size
- change: Show image and audio input markers on installed models (#139)
- change: Show the headphones marker on recommended models that take audio (#140)
- change: Show when a model fails to load
- change: Update the engine to b12110
- change: Use headphones instead of an ear for the audio input marker
