Rename rcli to wally

.agents/skills/rcli-architecture/SKILL.md → .agents/skills/wally-architecture/SKILL.md

@@ -1,16 +1,16 @@
 ---
-name: rcli-architecture
-description: Where RCLI logic belongs — command layering, proto as SOT, kit vs CLI ownership, Apple MLX host vs rcli-cxx. Use when adding a command, moving inference logic, or deciding whether a bug is SDK or CLI.
+name: wally-architecture
+description: Where Wally logic belongs — command layering, proto as SOT, kit vs CLI ownership, Apple MLX host vs wally-cxx. Use when adding a command, moving inference logic, or deciding whether a bug is SDK or CLI.
 ---
 
-# RCLI architecture
+# Wally architecture
 
-Repo: `RunanywhereAI/RCLI`. Product CLI named `rcli`. It consumes a **packaged
+Repo: `RunanywhereAI/RCLI`. Product CLI named `wally`. It consumes a **packaged
 C++ desktop kit** via `find_package(RunAnywhere)`. It does not
 `add_subdirectory` or FetchContent the SDK, and it does not compile llama.cpp /
 Sherpa / ONNX / MLX from source.
 
-Product version (`project(rcli VERSION …)` in `CMakeLists.txt`) is independent
+Product version (`project(wally VERSION …)` in `CMakeLists.txt`) is independent
 of the SDK kit pin in `cmake/sdk-pin.cmake`.
 
 ## Ownership
@@ -27,7 +27,7 @@ argv / flags / env
 The kit owns truth: models, backends, proto contracts, download, inference.
 The CLI renders and interacts. If a command is composing a multi-step bootstrap,
 hardcoding an engine name, or post-processing model output, that is a bug in the
-SDK — fix it there, then consume a new kit (**rcli-kit-pin**).
+SDK — fix it there, then consume a new kit (**wally-kit-pin**).
 
 ## Layering rules
 
@@ -56,14 +56,14 @@ a retired MetalRT / hardcoded catalog.
 
 ## Apple MLX host
 
-On Apple Silicon, `cmake --build` produces `build/rcli` (Swift host wrapping
-`rcli_run_main`). Users never run `rcli-cxx`; that name exists only so CMake
-cannot overwrite the product binary. Independent clones set `RCLI_SDK_SWIFT_PATH`
-to a runanywhere-sdks checkout (CI does this). Nested `EXTERNAL/RCLI` finds
-`../../Package.swift` automatically. Disable with `-DRCLI_APPLE_MLX_HOST=OFF`
+On Apple Silicon, `cmake --build` produces `build/wally` (Swift host wrapping
+`wally_run_main`). Users never run `wally-cxx`; that name exists only so CMake
+cannot overwrite the product binary. Independent clones set `WALLY_SDK_SWIFT_PATH`
+to a runanywhere-sdks checkout (CI does this). Nested `EXTERNAL/Wally` finds
+`../../Package.swift` automatically. Disable with `-DWALLY_APPLE_MLX_HOST=OFF`
 only for a C++-only compile loop.
 
-NeuRT image gen is `#if RCLI_HAS_NEURT` in `src/commands/cmd_image.cpp`, which
+NeuRT image gen is `#if WALLY_HAS_NEURT` in `src/commands/cmd_image.cpp`, which
 is true only when the NeuRT overlay is applied. Public bottles stay OSS.
 `--engine qhexrt` / `qnn` / `npu` / `hexagon` map to
 `INFERENCE_FRAMEWORK_QHEXRT`. Local HNPU trees are inferred from `v75`/`v79`/

.claude/skills/rcli-device-e2e/SKILL.md → .agents/skills/wally-device-e2e/SKILL.md

@@ -1,36 +1,36 @@
 ---
-name: rcli-device-e2e
-description: Run engine-agnostic rcli modality e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices. Use when adding overlay backends, proving LLM/STT/TTS/VLM/embed/diffusion on device, or when a PC only has one modality's bundles on disk.
+name: wally-device-e2e
+description: Run engine-agnostic wally modality e2e on Apple Neural Engine (NeuRT) and Snapdragon Hexagon NPU (QHexRT) devices. Use when adding overlay backends, proving LLM/STT/TTS/VLM/embed/diffusion on device, or when a PC only has one modality's bundles on disk.
 ---
 
-# RCLI device modality e2e
+# Wally device modality e2e
 
 Do not write per-engine tests. The harness is `scripts/e2e-modalities.sh`,
 called from `scripts/e2e.sh`. Keys are **primitives** (`llm`, `stt`, `tts`,
-`vlm`, `embed`, `image`, `vad`, `rerank`, `segment`). rcli picks the engine
+`vlm`, `embed`, `image`, `vad`, `rerank`, `segment`). wally picks the engine
 from catalog framework, local path, or plugin priority. `--engine` is an
-override (`RCLI_E2E_ENGINE`), never a required test input.
+override (`WALLY_E2E_ENGINE`), never a required test input.
 
 ## Run
 
 ```bash
 # Public CI (modelless): skip every modality
-bash scripts/e2e.sh /path/to/rcli
+bash scripts/e2e.sh /path/to/wally
 
 # Device: discover whatever is already on disk, then run each primitive
 export RUNANYWHERE_HOME=/path/to/home          # already-pulled OSS models
-export RCLI_E2E_MODEL_ROOTS=/path/to/hnpu:/path/to/coreml
-bash scripts/e2e-modalities.sh /path/to/rcli
+export WALLY_E2E_MODEL_ROOTS=/path/to/hnpu:/path/to/coreml
+bash scripts/e2e-modalities.sh /path/to/wally
 
 # Or pin one primitive (path or catalog id)
-RCLI_E2E_LLM=/path/to/lfm2_5_230m_HNPU \
-RCLI_E2E_STT=/path/to/whisper_base_HNPU \
-RCLI_E2E_TTS=/path/to/kitten_micro_0_8_HNPU \
-RCLI_E2E_EMBED=/path/to/embeddinggemma_300m_HNPU \
-  bash scripts/e2e-modalities.sh /path/to/rcli
+WALLY_E2E_LLM=/path/to/lfm2_5_230m_HNPU \
+WALLY_E2E_STT=/path/to/whisper_base_HNPU \
+WALLY_E2E_TTS=/path/to/kitten_micro_0_8_HNPU \
+WALLY_E2E_EMBED=/path/to/embeddinggemma_300m_HNPU \
+  bash scripts/e2e-modalities.sh /path/to/wally
 ```
 
-`RCLI_E2E_AUTO=1` pulls small OSS catalog defaults the **registered** backends
+`WALLY_E2E_AUTO=1` pulls small OSS catalog defaults the **registered** backends
 can run (`smollm2`, `whisper-tiny`, `piper`, `minilm`, `silero`, `mlx-qwen3`,
 …). Never enable AUTO in public CI.
 
@@ -49,20 +49,20 @@ copied for LLM smoke — not because the engine is LLM-only. Catalog ids:
 | VLM | `internvl3_5_1b` (~10 GB) | — |
 | Image | `cosmos3_edge_diffusion` | `sd15` |
 
-`rcli pull` of a Hugging Face **repo page** is HTML. Pass the expanded
+`wally pull` of a Hugging Face **repo page** is HTML. Pass the expanded
 directory to `-m`. Download `v81/*` only on Hexagon v81.
 
 Skip with a clear "no bundle" when the tree is missing. Fail only when a
 model was selected and the command failed.
 
 ## Overlay gotchas
 
-- Public bottles never list `neurt` / `qhexrt`. Rebuild product `rcli` against
-  an overlay kit (`RCLI_SDK_KIT` pointing at that prefix).
+- Public bottles never list `neurt` / `qhexrt`. Rebuild product `wally` against
+  an overlay kit (`WALLY_SDK_KIT` pointing at that prefix).
 - **QHexRT:** QAIRT **2.48** on Snapdragon X2 Elite / Hexagon v81.
   `ADSP_LIBRARY_PATH` must be the fully expanded
   `...\lib\hexagon-v81\unsigned` path. Nested `%QNN_SDK_ROOT%` in `cmd /c set`
-  does not expand. Copy `QnnHtp*.dll` next to `rcli.exe`. FastRPC ~90s then
+  does not expand. Copy `QnnHtp*.dll` next to `wally.exe`. FastRPC ~90s then
   user-driver fallback is normal. Use a `.bat`, not nested `cmd /c`.
 - **NeuRT image:** `--prompt` and `--out` required; `--steps 4` for smoke.
   Compiled zip, not the HF repo HTML. Tree needs `TextEncoder.mlmodelc` /
@@ -71,8 +71,8 @@ model was selected and the command failed.
   SDK inserts `mtmd_default_marker()`. An extra `<image>` makes
   `mtmd_tokenize` see 0 media markers. A tiny PNG can `bad_alloc` in
   SmolVLM2 after the 512×512 warmup; skip or pass a real photo via
-  `RCLI_E2E_VLM`.
+  `WALLY_E2E_VLM`.
 - **segment:** binary P6 PPM, not PNG.
 - STT has no `--engine` flag; put `-m` before the wav.
 
-See `rcli-e2e` for bottle/backends assertions and Apple MLX host link flags.
+See `wally-e2e` for bottle/backends assertions and Apple MLX host link flags.