From 6f3eabd14adceac1e071c2431548371d5c23ec5a Mon Sep 17 00:00:00 2001
From: Berry Wahlberg <40695099+BerryUIKI@users.noreply.github.com>
Date: Wed, 23 Sep 2026 19:43:04 +0800
Subject: [PATCH] docs: reconcile support matrix, roadmap, and acceptance
 dossier with 4-tier verification taxonomy

---
 docs/ACCEPTANCE_DOSSIER.md | 62 +++++++++++++++++++-------------------
 docs/ROADMAP_POST_V0_1.md  | 14 ++++-----
 docs/SUPPORT_MATRIX.md     | 55 +++++++++++++++++----------------
 3 files changed, 67 insertions(+), 64 deletions(-)

diff --git a/docs/ACCEPTANCE_DOSSIER.md b/docs/ACCEPTANCE_DOSSIER.md
index 2586f4a..dfc3937 100644
--- a/docs/ACCEPTANCE_DOSSIER.md
+++ b/docs/ACCEPTANCE_DOSSIER.md
@@ -100,38 +100,38 @@ cargo run --manifest-path launcher/Cargo.toml
 
 ## 5. Requirement-to-Evidence Matrix (R01–R18 & L01–L12)
 
-| Requirement ID | Specification | Implementation & Verification Evidence | Status |
+| Requirement ID | Specification | Implementation & Verification Evidence | Verification Status |
 | :--- | :--- | :--- | :--- |
-| **R01** | Lightweight startup | Backend boots without `torch`; cloud mode requires 0 local GPU. Verified by `test_health.py` and `test_m1_core.py`. | **VERIFIED** |
-| **R02** | Device readiness | `/api/v1/hardware/readiness` inspects NVIDIA GPU, VRAM, and storage without modifying system drivers. Verified by `test_m5_release.py`. | **VERIFIED** |
-| **R03** | Managed engine installation | Isolated virtualenvs in `~/.ai-workflow/engine/`. Interrupted install recorded in manifest. Verified by `test_m2_engines.py`. | **VERIFIED** |
-| **R04** | Existing engine connection | Connects to external engines via URL without modifying user directories. Verified by `test_comfy_bridge.py`. | **VERIFIED** |
-| **R05** | Engine lifecycle | Supervised Start, Stop, and Status. Probes HTTP responsiveness. External processes preserved. Verified by `test_m2_engines.py`. | **VERIFIED** |
-| **R06** | Infinite canvas | React Flow infinite canvas with pan/zoom, card selection, contextual image actions, and zero spaghetti wires. Verified by `FlowCanvas.tsx`. | **VERIFIED** |
-| **R07** | Image creation | Prompt, aspect ratio, seed, and model selection. Unsupported combinations explain what is missing. Verified by `test_cancellation_and_actions.py`. | **VERIFIED** |
-| **R08** | Inpainting | Canvas-aligned mask drawing tool (`InpaintModal.tsx`), natural dimension scaling, prompt editing, and result placement. Verified by `test_m3_creative.py`. | **VERIFIED** |
-| **R09** | Upscaling and export | 2×/4× resolution upscaling (`UpscaleModal.tsx`), thumbnail dimension preview, and local export/download (`ImageCardNode.tsx`). | **VERIFIED** |
-| **R10** | Workflow templates | Curated macro graphs for txt2img, img2img, inpaint, upscale. Node inputs validated. Verified by `test_macro_compiler.py`. | **VERIFIED** |
-| **R11** | Model inventory | Non-destructive scanning of safetensors/checkpoints. Detects architectures (SDXL, SD1.5, Flux, LoRA) and missing VAE/CLIP. Verified by `test_m5_release.py`. | **VERIFIED** |
-| **R12** | Directory management | Add/remove model roots via CLI and GUI without deleting files. Verified by `test_m5_release.py` and `EnvironmentManagerModal.tsx`. | **VERIFIED** |
-| **R13** | BYOK cloud | Local credential storage for OpenAI, Fal.ai, SiliconFlow. Secrets redacted and never saved to canvas or git. Verified by `test_m4_cloud.py`. | **VERIFIED** |
-| **R14** | Task lifecycle & cancellation | Queued, running, completed, error, and cancelled states. REST cancel endpoints (`/api/v1/tasks/{id}/cancel`, `/api/v1/workflow/cancel/{id}`) with truthful cloud disclosure. Verified by `test_cancellation_and_actions.py`. | **VERIFIED** |
-| **R15** | Persistence & asset adoption | SQLite database saves projects, assets, and canvas state. Remote cloud URLs adopted into local content-addressable storage. Verified by `test_m1_core.py` and `test_m4_cloud.py`. | **VERIFIED** |
-| **R16** | Deterministic caching | Hash calculated from action, prompt, model, seed, input image, and mask hashes. Identical inputs reuse cache; parameter changes re-execute. Verified by `test_dag_cache.py`. | **VERIFIED** |
-| **R17** | Portable architecture | OS-specific process flags behind adapters (`launcher/src/backend.rs`). Portable core tested with 73 automated tests. | **VERIFIED** |
-| **R18** | Native entry points | Opens configured native UI (`/api/v1/runtime/{engine}/ui`) without hardcoded assumptions. | **VERIFIED** |
-| **L01** | Single entry point | Windows release package requires zero developer tools. Verified by `smoke-test-package.ps1` with sanitized PATH. | **VERIFIED** |
-| **L02** | Accurate startup | 4-stage readiness check in Rust launcher. Missing frontend renders diagnostic HTML error rather than blank page. Verified by `test_m6_launcher_manager.py`. | **VERIFIED** |
-| **L03** | No duplicate processes | Windows Named Mutex (`Global\BerryAIStudioLauncherMutex`) focuses existing workspace on second launch. Port collisions detected with diagnostic guidance. | **VERIFIED** |
-| **L04** | Clear status | `/api/v1/manager/status` and `berry.exe status` aggregate core, engines, cloud, models, and active tasks. Verified by `test_m6_launcher_manager.py`. | **VERIFIED** |
-| **L05** | Controlled engines | Managed engines install, start, stop, and update independently. External engines are never stopped or modified. | **VERIFIED** |
-| **L06** | Recovery | Actionable error messages and retry paths for broken virtualenvs, port conflicts, and interrupted updates. | **VERIFIED** |
-| **L07** | Shutdown policy | `/api/v1/manager/shutdown` and `berry stop` check active tasks; non-forced exit returns HTTP 409 if generations are running. Verified by `test_cancellation_and_actions.py`. | **VERIFIED** |
-| **L08** | Updates separation | Application updates (`POST /api/v1/updates/app`) and engine updates (`POST /api/v1/runtime/{engine}/update`) are separate operations. Verified by `test_m6_launcher_manager.py`. | **VERIFIED** |
-| **L09** | Launcher portability | Windows-specific Named Mutex and process creation behind `#[cfg(windows)]`. Cross-platform HTTP/TCP core. | **VERIFIED** |
-| **L10** | Model inventory in launcher | `berry models list/roots/add-root/remove-root` CLI and GUI Environment Manager. Reuses backend model store. | **VERIFIED** |
-| **L11** | Rust launcher boundary | Launcher implemented in Rust (`launcher/`), interacting with backend via REST API (`/api/v1/manager/*`). | **VERIFIED** |
-| **L12** | Safe engine updates & rollback | Before engine update, active tasks are checked. Pre-update git commit captured; on failure, `git checkout <commit>` restores code. Preserves models and configs. Verified by `test_m6_launcher_manager.py`. | **VERIFIED** |
+| **R01** | Lightweight startup | Backend boots without `torch`; cloud mode requires 0 local GPU. Verified by `test_health.py` and `test_m1_core.py`. | **MOCK_VERIFIED** |
+| **R02** | Device readiness | `/api/v1/hardware/readiness` inspects NVIDIA GPU, VRAM, and storage without modifying system drivers. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **R03** | Managed engine installation | Isolated virtualenvs in `~/.ai-workflow/engine/`. Interrupted install recorded in manifest. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **R04** | Existing engine connection | Connects to external engines via URL without modifying user directories. Verified by `test_comfy_bridge.py`. | **MOCK_VERIFIED** |
+| **R05** | Engine lifecycle | Supervised Start, Stop, and Status. Probes HTTP responsiveness. External processes preserved. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **R06** | Infinite canvas | React Flow infinite canvas with pan/zoom, card selection, contextual image actions, and zero spaghetti wires. Verified in browser build. | **MOCK_VERIFIED** |
+| **R07** | Image creation | Prompt, aspect ratio, seed, and model selection. Real Fal.ai and local ComfyUI RTX 3060 runs recorded. | **REAL_PROVIDER_VERIFIED** & **PHYSICAL_VERIFIED** |
+| **R08** | Inpainting | Canvas-aligned mask drawing tool (`InpaintModal.tsx`), natural dimension scaling, prompt editing. Real Fal.ai run recorded. | **REAL_PROVIDER_VERIFIED** |
+| **R09** | Upscaling and export | 2×/4× resolution upscaling (`UpscaleModal.tsx`), thumbnail dimension preview, and local export/download (`ImageCardNode.tsx`). Real Fal.ai run recorded. | **REAL_PROVIDER_VERIFIED** |
+| **R10** | Workflow templates | Curated macro graphs for txt2img, img2img, inpaint, upscale. Node inputs validated. Verified by `test_macro_compiler.py`. | **MOCK_VERIFIED** |
+| **R11** | Model inventory | Non-destructive scanning of safetensors/checkpoints. Detects architectures (SDXL, SD1.5, Flux, LoRA) and missing VAE/CLIP. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **R12** | Directory management | Add/remove model roots via CLI and GUI without deleting files. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **R13** | BYOK cloud | Local credential storage for OpenAI, Fal.ai, SiliconFlow. Secrets redacted and never saved to canvas or git. Real BYOK credentials verified. | **REAL_PROVIDER_VERIFIED** |
+| **R14** | Task lifecycle & cancellation | Queued, running, completed, error, and cancelled states. REST cancel endpoints with truthful cloud disclosure. Verified by `test_cancellation_and_actions.py`. | **MOCK_VERIFIED** |
+| **R15** | Persistence & asset adoption | SQLite database saves projects, assets, and canvas state. Remote cloud URLs adopted into local content-addressable storage. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **R16** | Deterministic caching | Hash calculated from action, prompt, model, seed, input image, and mask hashes. Identical inputs reuse cache; parameter changes re-execute. Verified by `test_dag_cache.py`. | **MOCK_VERIFIED** |
+| **R17** | Portable architecture | OS-specific process flags behind adapters (`launcher/src/backend.rs`). Cross-platform code paths passing unit tests. | **SOURCE_COMPATIBLE** |
+| **R18** | Native entry points | Opens configured native UI (`/api/v1/runtime/{engine}/ui`) without hardcoded assumptions. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L01** | Single entry point | Windows release package requires zero developer tools. Verified by `smoke-test-package.ps1` with sanitized PATH. | **PHYSICAL_VERIFIED** |
+| **L02** | Accurate startup | 4-stage readiness check in Rust launcher. Missing frontend renders diagnostic HTML error rather than blank page. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L03** | No duplicate processes | Windows Named Mutex (`Global\BerryAIStudioLauncherMutex`) focuses existing workspace on second launch. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L04** | Clear status | `/api/v1/manager/status` and `berry.exe status` aggregate core, engines, cloud, models, and active tasks. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L05** | Controlled engines | Managed engines install, start, stop, and update independently. External engines are never stopped or modified. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L06** | Recovery | Actionable error messages and retry paths for broken virtualenvs, port conflicts, and interrupted updates. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L07** | Shutdown policy | `/api/v1/manager/shutdown` and `berry stop` check active tasks; non-forced exit returns HTTP 409 if generations are running. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L08** | Updates separation | Application updates (`POST /api/v1/updates/app`) and engine updates (`POST /api/v1/runtime/{engine}/update`) are separate operations. Verified by `test_m6_launcher_manager.py`. | **MOCK_VERIFIED** |
+| **L09** | Launcher portability | Windows-specific Named Mutex and process creation behind `#[cfg(windows)]`. Cross-platform HTTP/TCP core. | **SOURCE_COMPATIBLE** |
+| **L10** | Model inventory in launcher | `berry models list/roots/add-root/remove-root` CLI and GUI Environment Manager. Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L11** | Rust launcher boundary | Launcher implemented in Rust (`launcher/`), interacting with backend via REST API (`/api/v1/manager/*`). Verified on host rig. | **PHYSICAL_VERIFIED** |
+| **L12** | Safe engine updates & rollback | Before engine update, active tasks are checked. Pre-update git commit captured; on failure, `git checkout <commit>` restores code. Verified by `test_m6_launcher_manager.py`. | **MOCK_VERIFIED** |
 
 ---
 
diff --git a/docs/ROADMAP_POST_V0_1.md b/docs/ROADMAP_POST_V0_1.md
index b3a8a2b..5a28567 100644
--- a/docs/ROADMAP_POST_V0_1.md
+++ b/docs/ROADMAP_POST_V0_1.md
@@ -26,14 +26,14 @@ This roadmap operationalizes the long-term approved technical directions for Ber
 
 ## 2. Dependency-Ordered Milestones & Implementation Status
 
-| Milestone | Title & Focus Area | Implementation PR | Status on `dev` | Verification Evidence |
+| Milestone | Title & Focus Area | Implementation PRs | Status on `dev` | Verification Evidence & Tier |
 | :--- | :--- | :--- | :--- | :--- |
-| **M7** | Video Generation & Usable Video Creation Journey | [PR #11](https://github.com/BerryUIKI/AI-Studio/pull/11) | **Merged** (`64cd55c`) | `test_m7_video.py` (7/7 passed), SVD XT / AnimateDiff macro compilers, Fal / SiliconFlow cloud video, canvas video cards |
-| **M8** | Documented Unified Berry CLI for Creation & Management | [PR #12](https://github.com/BerryUIKI/AI-Studio/pull/12) | **Merged** (`8e394dd`) | `test_m8_cli.py` (3/3 passed), Rust tests (2/2 passed), `berry run/tasks/system`, [`docs/CLI_REFERENCE.md`](CLI_REFERENCE.md) |
-| **M9** | Conversational Agent Control of Existing Workflows & Engines | [PR #13](https://github.com/BerryUIKI/AI-Studio/pull/13) | **Merged** (`1260722`) | `test_m9_agent.py` (6/6 passed), `AgentService`, transparent `AgentProposal`, human-in-the-loop gate, frontend `AgentPanel` |
-| **M10** | Agent-Assisted Construction, Validation, Repair of Workflows | [PR #14](https://github.com/BerryUIKI/AI-Studio/pull/14) | **Merged** (`60c3620`) | `test_m10_workflow_repair.py` (6/6 passed), ComfyUI DAG syntax & port validator, automated heuristic repair engine |
-| **M11** | Local Inference on Selected Non-NVIDIA Discrete GPUs | [PR #15](https://github.com/BerryUIKI/AI-Studio/pull/15) | **Merged** (`7f700b2`) | `test_m11_hardware.py` (4/4 passed), AMD/Intel/Apple GPU detection, DirectML flags, [`docs/SUPPORT_MATRIX.md`](SUPPORT_MATRIX.md) |
-| **M12** | Cross-Platform Desktop Adapters (macOS Metal & Linux) | [PR #16](https://github.com/BerryUIKI/AI-Studio/pull/16) | **Merged** (`2551796`) | `test_m12_cross_platform.py` (4/4 passed), Unix domain socket single instance, cross-platform browser, [`docs/PLATFORM_PORTABILITY.md`](PLATFORM_PORTABILITY.md) |
+| **M7** | Video Generation & Usable Video Creation Journey | [PR #11](https://github.com/BerryUIKI/AI-Studio/pull/11), [PR #20](https://github.com/BerryUIKI/AI-Studio/pull/20) | **Merged** (`64cd55c`, `befb14d`) | **Mocked Verification**: `test_m7_video.py` (7/7 passed), SVD XT / AnimateDiff macro compiler, MIME detection, container vs. animated WebP handling. (Live cloud video generation requires funded BYOK keys). |
+| **M8** | Documented Unified Berry CLI for Creation & Management | [PR #12](https://github.com/BerryUIKI/AI-Studio/pull/12) | **Merged** (`8e394dd`) | **Physical Hardware Verified**: Rust CLI binary execution on Windows 11 host (`test_m8_cli.py`, `cargo test`, `berry.exe run/tasks/system`). |
+| **M9** | Conversational Agent Control of Existing Workflows & Engines | [PR #13](https://github.com/BerryUIKI/AI-Studio/pull/13), [PR #18](https://github.com/BerryUIKI/AI-Studio/pull/18) | **Merged** (`1260722`, `3cf860a`) | **Mocked Verification**: `test_m9_agent.py` (6/6 passed), `AgentService`, deterministic keyword/regex intent parsing, transparent `AgentProposal`, human-in-the-loop gate. |
+| **M10** | Agent-Assisted Construction, Validation, Repair of Workflows | [PR #14](https://github.com/BerryUIKI/AI-Studio/pull/14), [PR #18](https://github.com/BerryUIKI/AI-Studio/pull/18) | **Merged** (`60c3620`, `3cf860a`) | **Mocked Verification**: `test_m10_workflow_repair.py` (9/9 passed), ComfyUI DAG syntax & port validator, ambiguity guards, same-family model substitution blocker. |
+| **M11** | Local Inference on Selected Non-NVIDIA Discrete GPUs | [PR #15](https://github.com/BerryUIKI/AI-Studio/pull/15), [PR #19](https://github.com/BerryUIKI/AI-Studio/pull/19) | **Merged** (`7f700b2`, `c35f3b1`) | **Mocked Verification**: `test_m11_hardware.py` (5/5 passed), AMD/Intel GPU detection via WMI/ROCm, DirectML launch flags. (Physical AMD/Intel hardware runs unverified in this session). |
+| **M12** | Cross-Platform Desktop Adapters (macOS Metal & Linux) | [PR #16](https://github.com/BerryUIKI/AI-Studio/pull/16), [PR #19](https://github.com/BerryUIKI/AI-Studio/pull/19) | **Merged** (`2551796`, `c35f3b1`) | **Source Compatibility**: `test_m12_cross_platform.py` (4/4 passed), Unix domain socket single instance, cross-platform browser launchers. (Native packaging and macOS/Linux runtime binaries unverified). |
 
 ---
 
diff --git a/docs/SUPPORT_MATRIX.md b/docs/SUPPORT_MATRIX.md
index 7f39a19..8f4d626 100644
--- a/docs/SUPPORT_MATRIX.md
+++ b/docs/SUPPORT_MATRIX.md
@@ -4,49 +4,52 @@ This document defines the supported discrete GPU hardware, acceleration backends
 
 ---
 
-## 1. Discrete GPU Tier Classification
+## 1. Verification Status Taxonomy
 
-| Tier | Definition | Expected Performance | Verification Status |
-| :--- | :--- | :--- | :--- |
-| **Tier 1: Verified Reference Hardware (Physical Evidence)** | Tested and validated directly on physical hardware rigs with documented generation runs. Native acceleration with FP16 tensor core optimization. | High (SDXL < 5s, Flux < 15s) | **Physically Verified**: NVIDIA GeForce RTX (Ampere/Ada on Windows 11 with CUDA 12.x; verified on RTX 3060 12GB). |
-| **Tier 2: Candidate Reference Architecture (Unverified on Physical Hardware)** | Structural detection, launch flags, and DirectML / ROCm / IPEX adapters implemented and tested via mock fixtures. | Moderate (SD1.5 < 4s, SDXL < 15s) | **Candidate Architecture**: AMD Radeon (DirectML/ROCm), Intel Arc (DirectML/OneAPI IPEX). Physical generation unverified in this session. |
-| **Tier 3: Source-Compatible Platform (Unverified Native Package)** | Cross-platform code paths (Metal MPS, POSIX sockets, platform launchers) implemented and unit-tested. | Device dependent | **Source Compatible Only**: macOS Apple Silicon, Linux (Ubuntu). Native binary execution and packaging unverified in this session. |
-| **Tier 4: Cloud Recommended** | Systems with < 6GB VRAM, legacy architectures (pre-DirectX 12.1), integrated GPUs, or CPU-only setups. | Insufficient for local SDXL/Flux | GUI prompts user to configure Cloud BYOK (OpenAI, Fal.ai, SiliconFlow). |
+Berry AI Studio strictly distinguishes four distinct verification tiers across all documentation, tests, and user-facing reports:
+
+| Status | Exact Definition | Scope in Berry AI Studio |
+| :--- | :--- | :--- |
+| **Physical Hardware Verification** | Executed and confirmed on a physical host machine with documented hardware, drivers, and real generation outputs. | **Windows 11 (x86_64) with NVIDIA GeForce RTX 3060 12GB (Ampere)**, NVIDIA Driver 572.70, CUDA 12.8. Real ComfyUI local inference verified. |
+| **Mocked Verification** | Logic, detection, flags, and schemas verified via automated unit/integration tests using mocked I/O, simulated WMI controllers, or mock subprocess responses. | **AMD Radeon (DirectML/ROCm)**, **Intel Arc (DirectML/OneAPI IPEX)**, DAG cycle detection, hash caching, engine process supervisor state machines. |
+| **Source Compatibility** | Platform adapters, path normalization, browser spawn commands, and single-instance locks implemented and passing unit tests, without native packaging or physical execution on native OS machines. | **macOS (Apple Silicon Metal MPS)**, **Linux (Ubuntu 22.04+)**. Source code is portable; native runtime binaries and desktop execution unverified in this session. |
+| **Real Provider Verification** | Live network API requests executed against remote cloud infrastructure using authenticated user BYOK credentials. | **Fal.ai** (FLUX Schnell), **SiliconFlow**, and **OpenAI** (DALL-E) verified via real API calls. Advanced video models (Fal SVD, CogVideoX) require funded BYOK keys and are verified via contract schema checks. |
 
 ---
 
-## 2. Hardware Support Matrix
+## 2. Hardware Support & Verification Matrix
 
 ### NVIDIA GeForce & RTX (Native CUDA)
-| Hardware | Windows 10/11 | Linux (Ubuntu 22.04+) | Backend | Status |
+| Hardware | Windows 10/11 | Linux (Ubuntu 22.04+) | Backend | Verification Status |
 | :--- | :--- | :--- | :--- | :--- |
-| RTX 4090 / 4080 / 4070 (Ada) | Native CUDA | Native CUDA | `cuda` | Compatible (Candidate Architecture) |
-| RTX 3090 / 3080 / 3070 / 3060 (Ampere) | Native CUDA | Native CUDA | `cuda` | **Physically Verified on Windows Rig** (RTX 3060 12GB, driver 572.70) |
-| RTX 2080 / 2070 / 2060 (Turing) | Native CUDA | Native CUDA | `cuda` | Compatible (Candidate Architecture) |
-| GTX 1660 / 1650 (6GB) | Native CUDA | Native CUDA | `cuda` (lowvram) | Supported |
-| GTX 1080 / 1070 (Pascal) | Native CUDA | Native CUDA | `cuda` (fp32 fallback) | Supported |
+| RTX 3060 12GB (Ampere) | Native CUDA | Native CUDA | `cuda` | **Physical Hardware Verified** (Host rig: Driver 572.70, CUDA 12.8) |
+| RTX 4090 / 4080 / 4070 (Ada) | Native CUDA | Native CUDA | `cuda` | Mocked Verification (Candidate Architecture) |
+| RTX 3090 / 3080 / 3070 (Ampere) | Native CUDA | Native CUDA | `cuda` | Mocked Verification (Candidate Architecture) |
+| RTX 2080 / 2070 / 2060 (Turing) | Native CUDA | Native CUDA | `cuda` | Mocked Verification (Candidate Architecture) |
+| GTX 1660 / 1650 (6GB) | Native CUDA | Native CUDA | `cuda` (lowvram) | Mocked Verification (Candidate Architecture) |
+| GTX 1080 / 1070 (Pascal) | Native CUDA | Native CUDA | `cuda` (fp32 fallback) | Mocked Verification (Candidate Architecture) |
 
 ### AMD Radeon (DirectML & ROCm)
-| Hardware | Windows 10/11 | Linux (Ubuntu 22.04+) | Backend | Status |
+| Hardware | Windows 10/11 | Linux (Ubuntu 22.04+) | Backend | Verification Status |
 | :--- | :--- | :--- | :--- | :--- |
-| Radeon RX 7900 XTX / XT (RDNA3) | DirectML (`--directml`) | ROCm 6.1+ | `directml` / `rocm` | Candidate Reference Architecture (Physical execution unverified) |
-| Radeon RX 7800 XT / 7700 XT (RDNA3) | DirectML (`--directml`) | ROCm 6.1+ | `directml` / `rocm` | Candidate Reference Architecture (Physical execution unverified) |
-| Radeon RX 6800 XT / 6700 XT (RDNA2) | DirectML (`--directml`) | ROCm 5.7+ | `directml` / `rocm` | Candidate Reference Architecture (Physical execution unverified) |
-| Radeon RX 6600 / 6500 XT | DirectML (`--lowvram`) | Community ROCm | `directml` | Experimental |
+| Radeon RX 7900 XTX / XT (RDNA3) | DirectML (`--directml`) | ROCm 6.1+ | `directml` / `rocm` | Mocked Verification (Candidate Architecture; unverified physical hardware) |
+| Radeon RX 7800 XT / 7700 XT (RDNA3) | DirectML (`--directml`) | ROCm 6.1+ | `directml` / `rocm` | Mocked Verification (Candidate Architecture; unverified physical hardware) |
+| Radeon RX 6800 XT / 6700 XT (RDNA2) | DirectML (`--directml`) | ROCm 5.7+ | `directml` / `rocm` | Mocked Verification (Candidate Architecture; unverified physical hardware) |
+| Radeon RX 6600 / 6500 XT | DirectML (`--lowvram`) | Community ROCm | `directml` | Experimental (Unverified physical hardware) |
 | Older Radeon RX 5000 / Vega | DirectML | Not recommended | `directml` | Experimental / Cloud Recommended |
 
 ### Intel Arc Discrete GPUs
-| Hardware | Windows 10/11 | Linux (Ubuntu 22.04+) | Backend | Status |
+| Hardware | Windows 10/11 | Linux (Ubuntu 22.04+) | Backend | Verification Status |
 | :--- | :--- | :--- | :--- | :--- |
-| Intel Arc A770 (16GB) | DirectML (`--directml`) | OneAPI IPEX | `directml` / `ipex` | Candidate Reference Architecture (Physical execution unverified) |
-| Intel Arc A750 (8GB) | DirectML (`--directml`) | OneAPI IPEX | `directml` / `ipex` | Candidate Reference Architecture (Physical execution unverified) |
-| Intel Arc A580 / A380 | DirectML (`--lowvram`) | OneAPI IPEX | `directml` | Experimental |
+| Intel Arc A770 (16GB) | DirectML (`--directml`) | OneAPI IPEX | `directml` / `ipex` | Mocked Verification (Candidate Architecture; unverified physical hardware) |
+| Intel Arc A750 (8GB) | DirectML (`--directml`) | OneAPI IPEX | `directml` / `ipex` | Mocked Verification (Candidate Architecture; unverified physical hardware) |
+| Intel Arc A580 / A380 | DirectML (`--lowvram`) | OneAPI IPEX | `directml` | Experimental (Unverified physical hardware) |
 
 ### Apple Silicon (macOS Metal)
-| Hardware | macOS 14+ (Sonoma) | macOS 13 (Ventura) | Backend | Status |
+| Hardware | macOS 14+ (Sonoma) | macOS 13 (Ventura) | Backend | Verification Status |
 | :--- | :--- | :--- | :--- | :--- |
-| Apple M1 / M2 / M3 / M4 (Max / Pro) | Native Metal (MPS) | Native Metal (MPS) | `mps` | Source Compatible (Native build/run unverified) |
-| Apple M1 / M2 / M3 (Base 8GB/16GB) | Native Metal (`--lowvram`) | Native Metal (`--lowvram`) | `mps` | Source Compatible (Native build/run unverified) |
+| Apple M1 / M2 / M3 / M4 (Max / Pro) | Native Metal (MPS) | Native Metal (MPS) | `mps` | Source Compatibility (Native build/run unverified) |
+| Apple M1 / M2 / M3 (Base 8GB/16GB) | Native Metal (`--lowvram`) | Native Metal (`--lowvram`) | `mps` | Source Compatibility (Native build/run unverified) |
 
 ---