Limit serving engines to Lemonade and vLLM

.github/workflows/ci.yml

@@ -133,13 +133,9 @@ jobs:
       - name: Acceptance harness self-tests
         if: needs.changes.outputs.heavy == 'true'
         run: |
-          python scripts/pytorch_therock_gpu_test.py --self-test
-          python scripts/llama_cpp_therock_gpu_test.py --self-test
           python scripts/comfyui_therock_gpu_test.py --self-test
           python scripts/local_assistant_therock_gpu_test.py --self-test
           python scripts/vllm_therock_gpu_test.py --self-test
-          python scripts/sglang_therock_gpu_test.py --self-test
-          python scripts/atom_therock_gpu_test.py --self-test
           python scripts/wsl_preflight.py --self-test
 
       - name: Portable WSL build deps self-test
@@ -173,7 +169,7 @@ jobs:
         shell: pwsh
         run: |
           cargo build --workspace --all-targets
-          cargo build --release -p rocm -p rocmd -p rocm-engine-pytorch -p rocm-engine-llama-cpp -p rocm-engine-lemonade -p rocm-engine-atom -p rocm-engine-vllm -p rocm-engine-sglang -p xtask
+          cargo build --release -p rocm -p rocmd -p rocm-engine-lemonade -p rocm-engine-vllm -p xtask
 
       - name: Test
         if: needs.changes.outputs.heavy == 'true'
@@ -189,13 +185,9 @@ jobs:
         if: needs.changes.outputs.heavy == 'true'
         shell: pwsh
         run: |
-          python .\scripts\pytorch_therock_gpu_test.py --self-test
-          python .\scripts\llama_cpp_therock_gpu_test.py --self-test
           python .\scripts\comfyui_therock_gpu_test.py --self-test
           python .\scripts\local_assistant_therock_gpu_test.py --self-test
           python .\scripts\vllm_therock_gpu_test.py --self-test
-          python .\scripts\sglang_therock_gpu_test.py --self-test
-          python .\scripts\atom_therock_gpu_test.py --self-test
           python .\scripts\wsl_preflight.py --self-test
 
       - name: Release readiness self-test

AGENTS.md

@@ -128,13 +128,13 @@ Current workspace members:
 
 - apps: `apps/rocm`, `apps/rocmd`
 - shared crates: `crates/rocm-core`, `crates/rocm-engine-protocol`
-- engine crates: `engines/atom`, `engines/lemonade`, `engines/llama-cpp`, `engines/pytorch`, `engines/sglang`, `engines/vllm`
+- engine crates: `engines/lemonade`, `engines/vllm`
 
 Guardrails:
 
 - `crates/rocm-engine-protocol` is a contract surface; verify all impacted engines after protocol changes
 - preserve strict GPU-required behavior; do not introduce silent CPU fallback
-- respect platform gates (for example, native Windows handling for vLLM/SGLang)
+- respect platform gates (for example, native Windows handling for vLLM)
 - pin third-party GitHub Actions to a full commit SHA with a trailing `# vX.Y.Z` comment, never a moving tag (`@v2`, `@main`); a retagged or compromised action otherwise enters CI silently. Bump the SHA and comment together when upgrading
 - supported host platforms are Windows and Linux only (including WSL where documented)
 - platforms outside Windows/Linux are unsupported; do not implement, debug, or "fix" unsupported-platform behavior

Cargo.toml

@@ -9,11 +9,7 @@ members = [
   "crates/rocm-dash-collectors",
   "crates/rocm-dash-daemon",
   "crates/rocm-dash-tui",
-  "engines/llama-cpp",
-  "engines/atom",
-  "engines/pytorch",
   "engines/lemonade",
-  "engines/sglang",
   "engines/vllm",
   "xtask",
 ]

MANIFEST.md

@@ -577,7 +577,7 @@ from the workspace source. The user-facing binaries are:
   processes and expose a local HTTP API
 
 In addition, each engine crate under `engines/` builds its own
-`rocm-engine-<name>` host binary (for example `rocm-engine-pytorch`,
+`rocm-engine-<name>` host binary (for example `rocm-engine-lemonade`,
 `rocm-engine-vllm`), spawned by `rocmd` to run a specific inference engine.
 
 All binaries are compiled from the workspace source using the standard Cargo
@@ -629,13 +629,12 @@ distribution; no Python packages are installed for Lemonade by rocm-cli.
 
 ### Engine-Specific Python Dependencies
 
-The PyTorch engine manages its own Python virtual environment using the `uv`
-binary described above. Python packages are installed from the TheRock PyPI
-index and, where applicable, from public PyPI (`https://pypi.org`). No Python
-packages are bundled in the repository.
-
-The vLLM, SGLang, ATOM, and llama.cpp engines do not install Python packages
-automatically; they record externally-provided runtimes supplied by the user.
+The vLLM engine manages Python packages through the `uv` binary described
+above, installed from the TheRock PyPI index and, where applicable, from public
+PyPI (`https://pypi.org`). No Python packages are bundled in the repository. The
+vLLM engine can also record an externally-provided runtime supplied by the user
+instead of installing one. Lemonade ships as a self-contained embeddable
+distribution and installs no Python packages through rocm-cli.
 
 ## Repository Structure Notes
 

README.md

@@ -24,8 +24,8 @@ ROCm CLI is a command-line tool for setting up and running local AI on AMD GPUs,
 full-screen TUI dashboard for GPU telemetry, model serving, and chat.
 
 A single prebuilt binary for Linux and Windows. No Python, Rust, or existing
-ROCm install required. Ships with inference engine adapters for PyTorch,
-llama.cpp, Lemonade, ATOM, vLLM, and SGLang.
+ROCm install required. Ships with inference engine adapters for Lemonade and
+vLLM.
 
 > [!IMPORTANT]
 > **Tech Preview** -- This software is provided as-is, without warranty or
@@ -166,14 +166,14 @@ rocm engines install <engine> [--runtime-id KEY] [--python-version X.Y] [--reins
 rocm engines shell <engine>   [--runtime-id KEY | --env-id ID] [--shell PATH]
 ```
 
-Supported engines: `lemonade`, `pytorch`, `llama.cpp`, `atom`, `vllm`, `sglang`.
+Supported engines: `lemonade`, `vllm`.
 
 ### Model serving
 
 Start a local OpenAI-compatible model server:
 
 ```
-rocm serve <model> [--engine lemonade|pytorch|llama.cpp|atom|vllm|sglang]
+rocm serve <model> [--engine lemonade|vllm]
                    [--device gpu_required|gpu_preferred|cpu_only]
                    [--gpu auto|<index>]
                    [--runtime-id KEY | --env-id ID]
@@ -369,6 +369,4 @@ See `docs/commit-signatures.md` for details (GPG signing, GitHub "Verified", and
 - Testing and verification: `docs/testing.md`
 - Developer manual QA: `docs/manual-testing.md`
 - Engine plugin policy: `docs/engine-plugins.md`
-- ATOM adapter: `docs/atom.md`
 - vLLM adapter: `docs/vllm.md`
-- SGLang adapter: `docs/sglang.md`

apps/rocm/Cargo.toml

@@ -26,12 +26,8 @@ rocm-core = { path = "../../crates/rocm-core" }
 rocm-dash-daemon = { path = "../../crates/rocm-dash-daemon" }
 rocm-dash-tui = { path = "../../crates/rocm-dash-tui" }
 tokio = { workspace = true }
-rocm-engine-atom = { path = "../../engines/atom" }
 rocm-engine-lemonade = { path = "../../engines/lemonade" }
-rocm-engine-llama-cpp = { path = "../../engines/llama-cpp" }
-rocm-engine-pytorch = { path = "../../engines/pytorch" }
 rocm-engine-protocol = { path = "../../crates/rocm-engine-protocol" }
-rocm-engine-sglang = { path = "../../engines/sglang" }
 rocmd = { path = "../rocmd" }
 rocm-engine-vllm = { path = "../../engines/vllm" }
 rpassword.workspace = true