jundot/omlx
38.6
Weak · 17 November 2026
348.3k
lines of production code
Python
with Swift, JavaScript
5
measurements over time
What this system is
Features
Added maintainer scripts for benchmark corpora, FP16 model cloning, and ternary weight repacking
Three new Python scripts have been added to the repository to support model optimization and benchmarking workflows. \build\_benchmark\_corpora.py\ generates stable, pinned text corpora (including code, mixed languages, and literary works) for throughput benchmarking. \clone\_mlx\_model\_fp16.py\ creates FP16 clones of MLX quantized models to enable the optional Qwen3.5/3.8 ANE+CPU prefill path, allowing BNNS to consume FP16 activations directly. \repack\_ternary\_t5.py\ losslessly repacks MLX 2-bit ternary checkpoint weights into a more compact t5 (base-3) format, reducing weight size by approximately 20% while maintaining bit-identical dequantized values.
scripts · high confidence
Initial Homebrew formula for oMLX with macOS-specific build fixes
Users can now install the oMLX LLM inference server via Homebrew on Apple Silicon (macOS Sequoia and later). The formula installs the server and its Python dependencies, including optional support for native custom kernels (Bonsai, GLM-5.2, MiniMax M3, Qwen3.5/3.6/4) and structured output via xgrammar. It also bundles the spaCy English model for Kokoro TTS and includes a Homebrew service for running oMLX as a background daemon. The installation process includes specific workarounds for macOS 27 beta, such as building certain packages from source to avoid dylib stripping issues and patching xgrammar's library paths to ensure the Reasoning Parser works correctly.
Formula · high confidence
Introduce native macOS configuration and CLI management
The macOS app now includes a dedicated configuration layer that manages application settings, API keys, and CLI integration. Users benefit from a new \APIKeyGenerator\ that creates secure, crypto-grade API keys for authentication. The \AppConfig\ module serves as the single source of truth for settings (host, port, base path, model directories), ensuring that configuration changes like port updates are preserved across app restarts and that wildcard bind addresses are correctly normalized for client connections. Additionally, the \ShellEnvWriter\ introduces a managed CLI shim (\\~/.omlx/bin/omlx\) that allows the app to expose terminal commands without automatically modifying shell initialization files; instead, it prompts the user for consent before editing shell rc files, ensuring a safer and more transparent setup experience.
apps/omlx-mac/Sources/Config · high confidence
Native Swift in-app updater replaces Sparkle
The macOS application now uses a custom Swift-based updater instead of the Sparkle framework. This new system checks for updates via GitHub Releases, supporting Stable, Release Candidate, and Dev channels, and handles the download, mounting, and atomic swap of the application bundle using a launchd worker to ensure crash safety.
apps/omlx-mac/Sources/Updater · high confidence
Native Swift macOS app with comprehensive UI and menubar controls
The oMLX macOS application has been rewritten as a native Swift app, introducing a full suite of SwiftUI screens including Server, Status, Logs, Models, Downloads, Integrations, Security, Appearance, Performance, and benchmarking (Throughput, Accuracy, Context). The update adds a macOS menubar with live system stats and metric popovers, local usage history tracking, and a custom Swift-based updater (replacing Sparkle) that handles GitHub Releases and DMG cleanup. It also includes model card sheets, quantization settings, and native UI components for a consistent macOS experience.
apps/omlx-mac/oMLX.xcodeproj · high confidence
Native Swift server lifecycle and control for macOS
The macOS app now manages the backend server using a native Swift implementation instead of the previous Python-based manager. This introduces a Unix-domain socket control interface (AppControlServer) for sending start, stop, restart, and status commands, and a dedicated process lifecycle manager (ServerProcess) that handles health checks, automatic restarts with backoff, and graceful shutdowns. The new server also includes a port conflict resolver (PortConflictResolver) to detect and handle occupied ports, and a Python runtime locator (PythonRuntime) that correctly finds the bundled Python interpreter and configures environment variables like PYTHONHOME and PYTHONPATH for the macOS app bundle. Signal handlers (SignalHandlers) ensure the server process is cleaned up if the parent app is terminated unexpectedly.
apps/omlx-mac/Sources/Server · high confidence
Native macOS app gains comprehensive API data models
The native macOS app now includes a full suite of Swift Data Transfer Objects (DTOs) to communicate with the server's admin API, enabling features previously limited to the web dashboard. This change introduces structured models for benchmarking (throughput and accuracy), Hugging Face and ModelScope downloads and uploads, oQ quantization tasks, model settings (including experimental features like DFlash and MTP), profile management, and server statistics. By decoding these specific API responses, the app can now surface hardware telemetry, manage model presets, and expose advanced configuration options directly in the native interface. Also in this area: 6 commits (2 fixes) modifying apps/omlx-mac/Sources; GGUF model support, download file selection, and Claude Desktop integration; Fixes for Top-K validation and embedding context window settings.
apps/omlx-mac/Sources/Net/DTO · medium confidence · unverified
Native macOS app introduces dedicated screens for settings, benchmarks, and integrations
The macOS application now features a comprehensive set of native SwiftUI screens that replace or supplement the previous web-based interface. Users can configure appearance settings (such as menu bar refresh rates, dock icon visibility, and model library scope) directly in the app. New benchmark screens allow users to measure model accuracy and usable context windows, with results that can be exported. The Downloads screen supports model acquisition from both Hugging Face and ModelScope, including live progress tracking and model card previews. Integrations are managed natively, allowing users to route Claude Code requests to local or cloud models and configure other AI tools. Additionally, a new About screen provides version details, license information, and project links, while a Logs screen offers real-time server log monitoring with file selection and copy capabilities. Also in this area: 6 commits (2 fixes) modifying apps/omlx-mac/Sources; GGUF model support, download file selection, and Claude Desktop integration; Fixes for Top-K validation and embedding context window settings.
apps/omlx-mac/Sources/AppView/Screens · medium confidence · unverified
Native macOS app with menubar-first launch and controlled window lifecycle
The macOS application is now a native Swift app that launches primarily as a menubar status item, keeping the main Admin Panel window hidden until explicitly opened by the user or via the Welcome wizard on first run. The app manages its own lifecycle to preserve the background server and menubar even when the main window is closed, allowing users to quit the UI without stopping the server. It also includes a dedicated CLI shim for command-line integration and handles external URL schemes to target the main window to the correct application instance.
apps/omlx-mac/Sources/App · high confidence
Native macOS networking layer for the OMLX admin API
The macOS app now includes a dedicated Swift networking module (\Net\) that provides a typed, async HTTP client for the \/admin/api/\*\ surface. This change introduces \Endpoints.swift\, which centralizes all API path constants (covering models, profiles, Hugging Face downloads, benchmarks, and settings), and \OMLXClient.swift\, which implements the client with cookie-jar session management, automatic API-key authentication, and snake-case JSON serialization. This replaces ad-hoc networking code with a structured, reusable foundation for all macOS-specific API interactions. Also in this area: 6 commits (2 fixes) modifying apps/omlx-mac/Sources; GGUF model support, download file selection, and Claude Desktop integration.
apps/omlx-mac/Sources/Net · medium confidence · unverified
New first-run welcome wizard for macOS
A new first-run welcome wizard has been added to the macOS application, guiding users through a three-step flow: product introduction, configuration setup, and completion. This transient window appears only on the initial launch (when settings.json does not exist) and allows users to configure the base path, model directory, port, and API key before starting the server. Closing the wizard before completion cancels the setup without persisting configuration, ensuring a clean slate for the next launch.
apps/omlx-mac/Sources/Welcome · high confidence
New macOS admin screens for benchmarks, downloads, integrations, and logs
The macOS app now includes dedicated admin screens for running accuracy and context benchmarks, managing model downloads from Hugging Face and ModelScope, configuring third-party integrations (Claude Code, Codex, etc.), and viewing server logs. These new view models power the UI for these features, handling server communication, polling, and state management. Also in this area: 6 commits (2 fixes) modifying apps/omlx-mac/Sources; GGUF model support, download file selection, and Claude Desktop integration; Fixes for Top-K validation and embedding context window settings.
apps/omlx-mac/Sources/AppView/ViewModels · medium confidence · unverified
New macOS build and validation tooling for the native Swift app
The macOS packaging pipeline now includes a comprehensive build script (build.sh) that automates the creation of the native Swift oMLX.app, handling Swift compilation, venvstacks Python environment staging, and code signing. To ensure stability, a new Python script (check\_custom\_kernel\_python\_abi.py) validates that custom native kernels match the app's bundled Python ABI, preventing runtime crashes from mismatched extension modules. Additionally, a lightweight HTTP stub (dev\_server.py) is provided to help developers verify the Swift app's server process integration during local iteration.
apps/omlx-mac/Scripts · high confidence
Port macOS menubar to native Swift with live metrics and system stats
The macOS menubar is now implemented in native Swift, replacing the previous Python-based construction. This change introduces live inference activity indicators (Live, Average, All-Time) with sparkline graphs in popovers, a combined system stats item for CPU, GPU, and memory usage, and a visibility watcher that detects hidden icons and offers an Auto-Fix for macOS Tahoe. The menubar also now includes a models submenu with library scope filtering, a web dashboard shortcut, and improved responsiveness under server load.
apps/omlx-mac/Sources/Menubar · high confidence
macOS app introduces native theming, localized number formatting, and macOS 26 glass effects
The macOS app now uses a dedicated theme system that resolves colors dynamically from system settings, including a new Enhanced Readability toggle that increases text contrast and adjusts status colors. Number counts in the UI are now formatted locally, using 万/亿 for Chinese locales and K/M/B for others. Visual surfaces support a new glass effect that uses native macOS 26 (Tahoe) liquid glass where available, with a fallback for earlier versions, and sidebar icons use gradient-filled squircle shapes to match the web design.
apps/omlx-mac/Sources/Theme · high confidence
Removals
Removal of the native macOS menubar application
The native macOS menubar application (oMLX Menubar App) has been removed from the packaging directory. This deletes the Python/PyObjC implementation files (\\_\init\\.py\, \\\main\\_.py\, \app.py\, \config.py\, \preferences.py\, \server\_manager.py\, \welcome.py\) that previously provided a native interface for managing the oMLX server, including server start/stop controls, health monitoring, and configuration preferences.
_packaging/omlx\app · high confidence
Behavioural changes
1 commit (0 fixes) modifying apps/omlx-web/omlx\_web, apps/omlx-web/omlx\_web/templates/components, apps/omlx-web/src
A change to existing behaviour in apps/omlx-web/omlx\_web, apps/omlx-web/omlx\_web/templates/components, apps/omlx-web/src — 1 commit, 3 files: “refactor: split the web UI into apps/omlx-web and add serve --headless (\#4359)” (refs \#4359).
_apps/omlx-web/omlx\_web, apps/omlx-web/omlx\web/templates/components, apps/omlx-web/src · low confidence · unverified
106 commits (34 fixes) modifying apps/omlx-mac/Resources
A change to existing behaviour in apps/omlx-mac/Resources — 106 commits (34 fixs), 76 files: “feat(cluster): Prefill Accelerator helper for NVIDIA GPUs in Docker (\#5163)”; “feat: add managed server lifecycle controls”; “fix(mac): restore the menu bar icon from Appearance and write menubar.log”; and 103 more (refs \#5163, \#2028, \#5164, \#3829, \#3914, …).
apps/omlx-mac/Resources · medium confidence · unverified
7 commits (2 fixes) modifying apps/omlx-cuda-helper
A change to existing behaviour in apps/omlx-cuda-helper — 7 commits (2 fixs), 26 files: “fix(cuda-helper): leave the OS 16 GiB when an integrated GPU plans its layers”; “chore(cuda-helper): bump the helper to 0.1.1”; “perf(cuda-helper): DeepSeek V4.1 expert, attention and window cache kernels, helper 0.1.3 (\#5195)”; and 4 more (refs \#5195, \#5187).
apps/omlx-cuda-helper, apps/omlx-mac/Sources/AskAI · medium confidence · unverified
Documentation and build configuration updates
The project documentation has been significantly expanded with new README translations for French, Japanese, Korean, and Chinese, alongside updates to the English README to reflect new features like Vision-Language Models (VLM), tiered KV caching, and Homebrew service management. A new SECURITY.md policy has been added to define vulnerability reporting and trust assumptions. On the build side, setup.py now supports optional native custom kernels (e.g., for GLM-5.2, MiniMax M3) via a \--with-custom-kernel\ flag, and .gitignore has been updated to exclude generated build artifacts, Swift architecture notes, and local working files.
(repo-wide) · high confidence
Introduce native SwiftUI theme components for the macOS app
The macOS app now uses a new set of native SwiftUI components for its interface, replacing previous custom implementations. This includes themed buttons (primary, destructive, plain), a click-to-copy code chip, structured list groups with labeled rows, native-style popup pickers, progress bars, and standardized section headers and footers. These components ensure the app's controls share bezel metrics, fonts, and disabled/pressed treatments with native macOS system controls, providing a consistent and native-feeling user experience.
apps/omlx-mac/Sources/Theme/Components · high confidence
Native macOS settings interface with server lifecycle controls
The macOS app now features a native SwiftUI settings shell that replaces previous UI implementations, providing a sidebar-based navigation for Server, Models, Benchmarks, and General settings. This change introduces direct server lifecycle controls (start, stop, restart) and a unified Apply mechanism for storage and port changes, ensuring the server restarts only when configuration actually differs. It also adds an update confirmation sheet that requires user consent before downloading and relaunching, and integrates enhanced readability and system-style theming across the interface.
apps/omlx-mac/Sources/AppView · high confidence
Packaging pipeline migrated to Swift app with venvstacks layer generation
The packaging directory no longer builds the macOS application bundle or DMG directly; instead, it generates the venvstacks Python layers (runtime and framework) that the Swift macOS app embeds. The old PyObjC menubar app and DMG pipeline have been removed, and the Swift app under apps/omlx-mac is now the only macOS bundle. The build script now supports --venvstacks-only and --print-fingerprint to refresh the export, and the framework layer has been renamed from mlx-framework to mlx-base. The venvstacks configuration no longer lists package versions directly; instead, requirements are generated at build time from pyproject.toml. A new swap\_platform\_wheels function allows replacing mlx and mlx-metal with platform-specific wheels for different macOS targets. The minimum macOS version requirement is now 15.0+ due to MLX \>= 0.29.2 compatibility.
packaging · high confidence
Fixes
102 commits (54 fixes) fixing apps/omlx-web
A fix in apps/omlx-web — 102 commits (54 fixs), 216 files: “feat(web): store and apply custom theme token overrides (\#5038)”; “fix(i18n): translate the Chinese console values still in English (\#4457)”; “fix(web): show Applications copy buttons next to each command”; and 99 more (refs \#5038, \#4457, \#4723, \#4673, \#4826, …). Also in this area: 6 commits (3 fixes) modifying apps/omlx-web.
(repo-wide) · medium confidence · unverified
31 commits (19 fixes) fixing apps/omlx-web/omlx\_web/templates
A fix in apps/omlx-web/omlx\_web/templates — 31 commits (19 fixs), 4 files: “feat(web): store and apply custom theme token overrides (\#5038)”; “fix(web): refocus the API key field after a failed login”; “refactor: split the web UI into apps/omlx-web and add serve --headless (\#4359)”; and 28 more (refs \#5038, \#4359, \#4171, \#4795, \#4498, …).
_apps/omlx-web/omlx\web/templates · medium confidence · unverified
Remove TurboQuant and native MTP restrictions from the web UI
The web admin dashboard no longer hides or disables the TurboQuant and native MTP (Lightning MTP) feature toggles. These settings are now available for all models, allowing users to enable these optimizations directly from the UI without encountering previous interface restrictions.
omlx · high confidence
Test coverage
4 commits adding/updating tests in apps/omlx-web/tests/ui; 7 commits adding/updating tests in tests; Added Python tests for macOS app bundle CLI wrappers, kernel ABI guards, and settings wiring; Added integration tests for TTS streaming, cache consistency, and real-model validation; Added unit tests for macOS app core components; Expanded test coverage for core engine and cache components.
Dependencies
Upgrade MLX stack and align dependencies for v0.32.2 compatibility
The project now requires Python 3.11–3.13 and pins the core MLX framework to version 0.32.2, along with specific commits for mlx-lm, mlx-vlm, mlx-embeddings, and mlx-audio to ensure ABI compatibility with custom kernel extensions. Build requirements now include cmake and nanobind 2.15.0 to match the MLX runtime. Several dependencies have been updated or added, including huggingface-hub (\>=1.19.0), transformers (\>=5.14.0,\<5.18), fastapi (\>=0.108.0), and ddgs (9.16.0), while optional extras like grammar, mcp, cluster, modelscope, audio, and paroquant have been refined with specific version pins and Python version markers.
(dependencies) · high confidence
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 37.2 → 38.6 (+1.4)
- Rubric changed (rubric-2026.11.2 → rubric-2026.11.4) — scores are not directly comparable.
Lenses
- Code Health 23.8 → 23.9 (+0.1)
- Architecture 90.6 → 91.4 (+0.8)
- Maturity 63.2 → 70.3 (+7.1)
- Readiness 36.5 → 38.2 (+1.7)
- Security 63.5 → 72.0 (+8.5)
- Domain Modelling 100.0 → 100.0 (0.0)
- Accessibility 42.5 → 43.0 (+0.5)
- Performance 100.0 → 100.0 (0.0)
Resolved (149)
- Change-coupling hub: dashboard.js → ModelSettingsScreen.swift, PerformanceScreen.swift, ModelSettingsScreenVM.swift, GlobalSettingsDTO.swift, ModelsDTO.swift, OMLXClient.swift, accuracy_benchmark.py, routes.py, config.py, settings.py (apps/omlx-web/omlx_web/static/js/dashboard.js)
- Duplicated block (10 lines × 13) (apps/omlx-mac/Sources/AppView/Screens/AppearanceScreen.swift)
- Duplicated block (10 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/AskAIScreen.swift)
- Duplicated block (10 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/ContextBenchScreen.swift)
- Duplicated block (10 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/IntegrationsScreen.swift)
- Duplicated block (10 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/SecurityScreen.swift)
- Duplicated block (10 lines × 2) (omlx/cluster/identity.py)
- Duplicated block (10 lines × 3) (apps/omlx-mac/Sources/AppView/Screens/AppearanceScreen.swift)
- Duplicated block (10 lines × 4) (apps/omlx-mac/Sources/AppView/Screens/NetworkScreen.swift)
- Duplicated block (10 lines × 9) (apps/omlx-mac/Sources/AppView/Screens/ModelSettingsScreen.swift)
- Duplicated block (10–11 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/IntegrationsScreen.swift)
- Duplicated block (10–12 lines × 3) (omlx/cluster/incidents.py)
- Duplicated block (10–15 lines × 3) (apps/omlx-mac/Sources/AppView/Screens/AppearanceScreen.swift)
- Duplicated block (11 lines × 5) (apps/omlx-mac/Sources/AppView/Screens/AppearanceScreen.swift)
- Duplicated block (11–14 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/DownloadsScreen.swift)
- Duplicated block (11–16 lines × 4) (apps/omlx-mac/Sources/AppView/Screens/AppearanceScreen.swift)
- Duplicated block (12 lines × 2) (apps/omlx-mac/Sources/AppView/AppView.swift)
- Duplicated block (12 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/ModelsScreen.swift)
- Duplicated block (12 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/ModelsScreen.swift)
- Duplicated block (12 lines × 2) (apps/omlx-mac/Sources/AppView/Screens/QuantizationScreen.swift)
- …and 129 more
New (428)
- Attention.call (cognitive 16) (apps/omlx-cuda-helper/omlx_cuda_helper/families/qwen4_exp/attention.py)
- Change-coupling hub: dashboard.js → ModelSettingsScreen.swift, PerformanceScreen.swift, ModelSettingsScreenVM.swift, ModelsDTO.swift, accuracy_benchmark.py, routes.py, config.py, settings.py (apps/omlx-web/omlx_web/static/js/dashboard.js)
- ClassTooLong: IntegrationsScreenVM (apps/omlx-mac/Sources/AppView/ViewModels/IntegrationsScreenVM.swift)
- Config.validate (cognitive 17) (apps/omlx-cuda-helper/omlx_cuda_helper/families/deepseek_v41/config.py)
- ControlServer._handle (cognitive 16) (apps/omlx-cuda-helper/omlx_cuda_helper/control.py)
- ControlServer._handle (cyclomatic 16) (apps/omlx-cuda-helper/omlx_cuda_helper/control.py)
- Duplicated block (10 lines × 2) (omlx/admin/routes.py)
- Duplicated block (10 lines × 2) (omlx/patches/inkling2/model.py)
- Duplicated block (10–18 lines × 2) (apps/omlx-mac/Sources/Updater/AppUpdater.swift)
- Duplicated block (11 lines × 2) (apps/omlx-cuda-helper/omlx_cuda_helper/wire.py)
- Duplicated block (11 lines × 2) (omlx/engine/batched.py)
- Duplicated block (11 lines × 2) (omlx/patches/inkling2/loader.py)
- Duplicated block (11 lines × 2) (omlx/patches/inkling2/oq_virtual.py)
- Duplicated block (12 lines × 2) (apps/omlx-cuda-helper/omlx_cuda_helper/families/deepseek_v41/ngram.py)
- Duplicated block (12 lines × 2) (apps/omlx-cuda-helper/omlx_cuda_helper/worker.py)
- Duplicated block (13 lines × 2) (apps/omlx-cuda-helper/omlx_cuda_helper/families/glm5_next/model.py)
- Duplicated block (13 lines × 2) (apps/omlx-cuda-helper/omlx_cuda_helper/families/qwen3_5/stage.py)
- Duplicated block (13 lines × 2) (apps/omlx-cuda-helper/omlx_cuda_helper/wire.py)
- Duplicated block (13 lines × 2) (omlx/patches/inkling2/dspark.py)
- Duplicated block (13 lines × 3) (omlx/memory_monitor.py)
- …and 408 more
Changes since last survey
- 118 commits — 56 feature/other, 62 fixes
By area
- apps/omlx-web — 22 commits
- apps/omlx-mac — 12 commits
- omlx/patches — 12 commits
- apps/omlx-cuda-helper — 11 commits
- omlx/admin — 11 commits
- omlx/api — 8 commits
- omlx/engine — 7 commits
- (root) — 3 commits
- omlx/affine4.py — 3 commits
- .github/workflows — 2 commits
- Formula/omlx.rb — 2 commits
- omlx/adapter — 2 commits
- omlx/cluster — 2 commits
- omlx/engine_core.py — 2 commits
- omlx/integrations — 2 commits
- omlx/prefill_helper — 2 commits
- omlx/scheduler.py — 2 commits
- docs/Splash.md — 1 commit
- docs/TESTING.md — 1 commit
- omlx/_version.py — 1 commit
Notable commits
- fix: fix(adapter): keep model text from forging harmony log lines (#5125)
- fix: fix(admin): check drafts when a profile is applied (#5105)
- fix: fix(admin): name the refused field when a settings save is rejected (#4513)
- fix: fix(admin): report MTP as unused when an external engine serves the model (#5168)
- fix: fix(admin): show recent decode speed per request, not the lifetime average (#5075)
- fix: fix(admin): stop a parked draft from blocking settings and profile saves (#4401)
- fix: fix(affine): keep attention sinks on the fused bf16 prefill path (#5088)
- fix: fix(api): keep the usual key order in tool schemas (#5161)
- fix: fix(api): surface a coded error when a tool_calls finish has no calls (#5139)
- fix: fix(bench): grow a short saved Splash context for the run (#4938)
- fix: fix(benchmark): batch tests wider than max concurrent requests report inflated tg TPS (#5312)
- fix: fix(benchmark): include gen_tps in the pp result log line (#5126)
- fix: fix(benchmark): pin Prefill First for benchmark runs
- fix: fix(brew): install the cluster extra so the Cluster tab finds NVIDIA prefill helpers
- fix: fix(cluster): name the link, not the head counts, when the link gate blocks TP (#3024)
- fix: fix(cluster): pass DeepSeek V4.1 tool arguments to mlx-lm as objects
- fix: fix(codex): skip the reverse DNS lookup in the instructions probe server
- fix: fix(cuda-helper): keep the bf16 roundings of Flash-Next kernels on sm_90 and newer
- fix: fix(cuda-helper): leave the OS 16 GiB when an integrated GPU plans its layers
- fix: fix(dflash): restore drafter context on exact prefix hits
- …and 98 more
Architecture
- Unchanged — 0 containers · 1 contexts · 0 edges
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
jundot/omlx was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 17 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit bf28cd6c43541ececeec01ba797377c913da5f8f — the exact code this score is about.
- Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.84+preprod-d645ca9bbedd.