diff --git a/Makefile b/Makefile
index be56103..e86c541 100644
--- a/Makefile
+++ b/Makefile
@@ -2,11 +2,14 @@ PYTHON ?= python3
 REPO_ROOT := $(CURDIR)
 export PYTHONPATH := $(REPO_ROOT)/python:$(REPO_ROOT)/compiler$(if $(PYTHONPATH),:$(PYTHONPATH))
 
-.PHONY: check examples native-sdk-check package package-test python-check repository-check shell-check test validate
+.PHONY: check examples native-sdk-check package package-test python-check repository-check shell-check static-check test validate
 
 python-check:
 	$(PYTHON) -m compileall -q compiler/netra_compiler python/netra_kernel tools/compiler tools/ci tests/compiler tests/package
 
+static-check:
+	$(PYTHON) -m ruff check compiler/netra_compiler python/netra_kernel tools tests --select F401,F811,F821,F841
+
 test:
 	$(PYTHON) -m unittest discover -s tests/compiler -v
 
@@ -29,7 +32,7 @@ shell-check:
 	git ls-files -z '*.sh' | xargs -0 -r -n1 bash -n
 	git diff --check
 
-check: python-check test package-test repository-check native-sdk-check validate examples shell-check
+check: python-check static-check test package-test repository-check native-sdk-check validate examples shell-check
 
 package:
 	PYTHON=$(PYTHON) bash tools/ci/build_package.sh
diff --git a/README.md b/README.md
index 7389432..62a17ef 100644
--- a/README.md
+++ b/README.md
@@ -110,6 +110,44 @@ stream. Direct fixed launches remain available when graph capture is disabled.
 The installed SDK exports `libnetra_engine.so.1`, `<netra/engine.h>`, CMake and
 pkg-config metadata, plus a framework-neutral `netra_kernel.Engine` binding.
 
+## Repository layout
+
+The public integration surface is separated from target-specific kernels and
+inference-engine adapters:
+
+```text
+netra-kernel/
+├── compiler/netra_compiler/   # AOT compiler, IR, planners, frontends, backends
+│   ├── backends/gfx950/       # Target-specific compiler implementation
+│   └── library_data/          # Packaged kernels, manifests, and schemas
+├── python/netra_kernel/       # Bundle, runtime, CLI, and adapter APIs
+│   ├── runtime/               # Framework-neutral Python FFI
+│   ├── sglang/                # SGLang adapter and startup plugin only
+│   └── assets/                # Deployment profiles and packaged overlays
+├── include/netra/engine.h     # Stable framework-neutral C ABI
+├── runtime/                   # Native runtime implementations
+├── examples/reference_engine/ # Minimal C consumer with no framework dependency
+├── kernels/                   # Architecture-specific kernel sources
+├── manifests/                 # Model, deployment, and tactic contracts
+├── schemas/                   # Public JSON schemas
+├── docker/sglang/             # Thin-plugin and accepted-source images
+├── tools/                     # Build, compiler, benchmark, and CI tooling
+├── tests/                     # Compiler, package, and GPU validation
+└── docs/                      # Integration guides and acceptance evidence
+```
+
+`netra-compiler` and `netra-kernel` are separate installable distributions.
+The compiler emits a deterministic engine directory; `libnetra_engine.so.1`
+loads and launches it; an optional adapter translates only the surrounding
+inference-engine boundary. New engines integrate through the
+`netra.inference_engines` entry-point group without importing SGLang code.
+
+The top-level `kernels/gfx950`, `manifests`, and `schemas` paths point to the
+same canonical data packaged in the wheels. This preserves convenient
+source-tree commands without maintaining duplicate catalogs that can drift.
+Architecture-specific harnesses, profiling scripts, and historical results
+remain outside the public SDK surface.
+
 ## Supported today
 
 | Capability | Current support |
diff --git a/compiler/netra_compiler/engine.py b/compiler/netra_compiler/engine.py
index 9d84783..6047d78 100644
--- a/compiler/netra_compiler/engine.py
+++ b/compiler/netra_compiler/engine.py
@@ -1,7 +1,6 @@
 from __future__ import annotations
 
 import hashlib
-import json
 import shutil
 from collections.abc import Mapping
 from pathlib import Path
@@ -16,7 +15,7 @@
 from .memory import BufferRequest, checked_mul, plan_buffers
 from .planner import Plan, plan_graph
 from .profiles import ShapeProfile
-from .types import canonical_json, stable_hash, write_json
+from .types import stable_hash, write_json
 
 
 FORMAT_VERSION = "netra-engine-1"
diff --git a/compiler/netra_compiler/tactics.py b/compiler/netra_compiler/tactics.py
index d668aef..4559759 100644
--- a/compiler/netra_compiler/tactics.py
+++ b/compiler/netra_compiler/tactics.py
@@ -2,7 +2,6 @@
 
 from collections.abc import Mapping
 from dataclasses import dataclass
-from pathlib import Path
 from typing import Any, Iterable
 
 from .contracts import GoldenKernelContract, KernelArgument, KernelContract, Launch, Workspace
diff --git a/docs/notes/gfx1151-causal-conv-stream64-2026-07-29.md b/docs/notes/gfx1151-causal-conv-stream64-2026-07-29.md
index 274368b..01f843b 100644
--- a/docs/notes/gfx1151-causal-conv-stream64-2026-07-29.md
+++ b/docs/notes/gfx1151-causal-conv-stream64-2026-07-29.md
@@ -37,7 +37,7 @@ The raw path delivered the expected ~1.07 s host benefit, but pair-b changed its
 The retained raw ASM and harness are:
 
 - `kernels/gfx1151/gdn/experiments/causal_conv1d_stream64_gfx1151.s`
-- `kernels/gfx1151/gdn/experiments/causal_conv1d_state_update_gfx1151.s`
+- `kernels/gfx1151/gdn/causal_conv1d_state_update_gfx1151.s`
 - `harness/gfx1151/gdn/causal_conv1d_stream64_launcher.hip`
 - `tools/benchmark/benchmark_causal_conv1d_stream64.py`
 
diff --git a/kernels/gfx1151/gdn/experiments/causal_conv1d_state_update_gfx1151.s b/kernels/gfx1151/gdn/experiments/causal_conv1d_state_update_gfx1151.s
deleted file mode 100644
index 0c6af3b..0000000
--- a/kernels/gfx1151/gdn/experiments/causal_conv1d_state_update_gfx1151.s
+++ /dev/null
@@ -1,92 +0,0 @@
-// SPDX-License-Identifier: MIT
-// Raw gfx1151 state writeback for T=8192,D=8192,W=4 BF16 causal convolution.
-.amdgcn_target "amdgcn-amd-amdhsa--gfx1151"
-.amdhsa_code_object_version 6
-.text
-.protected causal_conv1d_state_update_gfx1151
-.globl causal_conv1d_state_update_gfx1151
-.p2align 8
-.type causal_conv1d_state_update_gfx1151,@function
-causal_conv1d_state_update_gfx1151:
-s_load_b128 s[4:7], s[0:1], 0
-s_load_b64 s[8:9], s[0:1], 16
-s_waitcnt lgkmcnt(0)
-v_mov_b32_e32 v8, 0
-global_load_dword v9, v8, s[8:9]
-s_lshl_b32 s10, s2, 8
-v_lshlrev_b32_e32 v1, 1, v0
-v_add_nc_u32_e32 v1, s10, v1
-v_add_nc_u32_e32 v2, 0x07ff4000, v1
-global_load_ushort v3, v2, s[4:5]
-v_add_nc_u32_e32 v2, 16384, v2
-global_load_ushort v4, v2, s[4:5]
-v_add_nc_u32_e32 v2, 16384, v2
-global_load_ushort v5, v2, s[4:5]
-s_waitcnt vmcnt(0)
-v_readfirstlane_b32 s11, v9
-s_mul_i32 s11, s11, 49152
-s_add_u32 s6, s6, s11
-s_addc_u32 s7, s7, 0
-s_mul_i32 s12, s2, 768
-s_waitcnt_depctr 0
-v_mul_lo_u32 v6, v0, 6
-v_add_nc_u32_e32 v6, s12, v6
-global_store_short v6, v3, s[6:7]
-global_store_short v6, v4, s[6:7] offset:2
-global_store_short v6, v5, s[6:7] offset:4
-s_endpgm
-.section .rodata,"a",@progbits
-.p2align 6,0
-.amdhsa_kernel causal_conv1d_state_update_gfx1151
-.amdhsa_group_segment_fixed_size 0
-.amdhsa_private_segment_fixed_size 0
-.amdhsa_kernarg_size 24
-.amdhsa_user_sgpr_count 2
-.amdhsa_user_sgpr_kernarg_segment_ptr 1
-.amdhsa_wavefront_size32 1
-.amdhsa_enable_private_segment 0
-.amdhsa_system_sgpr_workgroup_id_x 1
-.amdhsa_system_sgpr_workgroup_id_y 0
-.amdhsa_system_sgpr_workgroup_id_z 0
-.amdhsa_system_vgpr_workitem_id 0
-.amdhsa_next_free_vgpr 10
-.amdhsa_next_free_sgpr 13
-.amdhsa_reserve_vcc 1
-.amdhsa_float_denorm_mode_32 3
-.amdhsa_float_denorm_mode_16_64 3
-.amdhsa_dx10_clamp 1
-.amdhsa_ieee_mode 1
-.amdhsa_workgroup_processor_mode 1
-.amdhsa_memory_ordered 1
-.amdhsa_forward_progress 1
-.end_amdhsa_kernel
-.text
-.Lend: .size causal_conv1d_state_update_gfx1151,.Lend-causal_conv1d_state_update_gfx1151
-.amdgpu_metadata
----
-amdhsa.kernels:
-  - .args:
-      - { .name: x, .offset: 0, .size: 8, .value_kind: global_buffer, .address_space: global, .actual_access: read_only }
-      - { .name: state, .offset: 8, .size: 8, .value_kind: global_buffer, .address_space: global, .actual_access: write_only }
-      - { .name: cache_index, .offset: 16, .size: 8, .value_kind: global_buffer, .address_space: global, .actual_access: read_only }
-    .group_segment_fixed_size: 0
-    .kernarg_segment_align: 8
-    .kernarg_segment_size: 24
-    .language: OpenCL C
-    .language_version: [2,0]
-    .max_flat_workgroup_size: 128
-    .name: causal_conv1d_state_update_gfx1151
-    .private_segment_fixed_size: 0
-    .sgpr_count: 15
-    .sgpr_spill_count: 0
-    .symbol: causal_conv1d_state_update_gfx1151.kd
-    .uniform_work_group_size: 1
-    .uses_dynamic_stack: false
-    .vgpr_count: 10
-    .vgpr_spill_count: 0
-    .wavefront_size: 32
-    .workgroup_processor_mode: 1
-amdhsa.target: amdgcn-amd-amdhsa--gfx1151
-amdhsa.version: [1,2]
-...
-.end_amdgpu_metadata
diff --git a/netra_compiler/__init__.py b/netra_compiler/__init__.py
deleted file mode 100644
index e25120b..0000000
--- a/netra_compiler/__init__.py
+++ /dev/null
@@ -1,14 +0,0 @@
-"""Source-tree import shim for ``python -m netra_compiler.cli``.
-
-Installed packages use ``compiler/pyproject.toml`` directly.
-"""
-
-from pathlib import Path
-
-__path__.append(str(Path(__file__).resolve().parents[1] / "compiler" / "netra_compiler"))
-
-from .contracts import KernelContract
-from .engine import compile_engine
-
-__all__ = ["KernelContract", "compile_engine"]
-__version__ = "0.2.0"
diff --git a/pyproject.toml b/pyproject.toml
index d346721..df3eee7 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -18,7 +18,7 @@ classifiers = [
 
 [project.optional-dependencies]
 sglang = ["torch>=2.8"]
-test = ["pytest>=8", "build>=1.2"]
+test = ["pytest>=8", "build>=1.2", "ruff>=0.16"]
 
 [project.scripts]
 netra-kernel = "netra_kernel.kernel_cli:main"
diff --git a/python/netra_kernel/cli.py b/python/netra_kernel/cli.py
index f14046c..1f36dbb 100644
--- a/python/netra_kernel/cli.py
+++ b/python/netra_kernel/cli.py
@@ -77,7 +77,6 @@ def _doctor(bundle_path: Path | None) -> dict:
     expected_source_snapshot = None
     bundle_result = None
     if bundle_path is not None:
-        payload = load_bundle(bundle_path)
         integration = integration_configuration(bundle_path, "sglang")
         required = tuple(integration.get("required_modules", ()))
         expected_source_snapshot = integration.get("source_snapshot")
diff --git a/python/netra_kernel/native.py b/python/netra_kernel/native.py
index cf6adbb..fd95967 100644
--- a/python/netra_kernel/native.py
+++ b/python/netra_kernel/native.py
@@ -10,7 +10,7 @@
 import ctypes
 import ctypes.util
 import os
-from importlib.resources import as_file, files
+from importlib.resources import files
 from pathlib import Path
 from typing import Iterator
 
diff --git a/tests/compiler/test_engine_manifest.py b/tests/compiler/test_engine_manifest.py
index ebcfe7b..5206ec1 100644
--- a/tests/compiler/test_engine_manifest.py
+++ b/tests/compiler/test_engine_manifest.py
@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-import hashlib
 import json
 import tempfile
 import unittest
diff --git a/tests/gfx1151/test_netra_mxfp4_sgl_decode.py b/tests/gfx1151/test_netra_mxfp4_sgl_decode.py
index 72c68cf..d73b015 100644
--- a/tests/gfx1151/test_netra_mxfp4_sgl_decode.py
+++ b/tests/gfx1151/test_netra_mxfp4_sgl_decode.py
@@ -3,7 +3,6 @@
 
 import ctypes
 import json
-import math
 import os
 from pathlib import Path
 
diff --git a/tests/gfx1151/test_netra_mxfp4_sgl_linear.py b/tests/gfx1151/test_netra_mxfp4_sgl_linear.py
index b4829b4..102e7db 100644
--- a/tests/gfx1151/test_netra_mxfp4_sgl_linear.py
+++ b/tests/gfx1151/test_netra_mxfp4_sgl_linear.py
@@ -93,12 +93,12 @@ def main() -> None:
     if status:
         raise RuntimeError(f"launch failed: {status}")
     torch.cuda.synchronize()
+    output = output_groups.view(-1, n)[:m].to(torch.bfloat16)
     if os.environ.get("NETRA_ARGS_DEBUG") == "1":
         print(output.view(torch.int32).cpu()[0, :8].tolist())
         return
 
     weight64 = dequant(packed_cpu, scales_cpu)
-    output = output_groups.view(-1, n)[:m].to(torch.bfloat16)
     reference = activation.cpu().to(torch.float64) @ weight64.t()
     ref_bf16 = reference.to(torch.bfloat16)
     got = output.cpu()
diff --git a/tools/benchmark/benchmark_extend_attention_tiles.py b/tools/benchmark/benchmark_extend_attention_tiles.py
index 400ef57..4cfb585 100755
--- a/tools/benchmark/benchmark_extend_attention_tiles.py
+++ b/tools/benchmark/benchmark_extend_attention_tiles.py
@@ -96,8 +96,11 @@ def call() -> None:
                 "max_abs_vs_64x64": max_abs,
                 "normalized_l2_vs_64x64": normalized_l2,
             })
-        del q, k, v, k_buffer, v_buffer, reference
-
+        # Drop the closure before its captured tensors so the next prefix does
+        # not temporarily retain two full attention fixtures.
+        call = output = None
+        q = k = v = k_buffer = v_buffer = reference = None
+        torch.cuda.empty_cache()
     module._get_block_sizes_for_extend_attention = original_blocks
     result = {
         "target": "gfx1151",
diff --git a/tools/benchmark/benchmark_qkvzba_split_copy.py b/tools/benchmark/benchmark_qkvzba_split_copy.py
index 4b678f9..a56f264 100755
--- a/tools/benchmark/benchmark_qkvzba_split_copy.py
+++ b/tools/benchmark/benchmark_qkvzba_split_copy.py
@@ -154,7 +154,10 @@ def launch_triton() -> None:
                 "bit_mismatches_qkv_z_b_a": mismatches,
             }
         )
-        del raw_tensors, triton_tensors
+        # The launch closures capture the tensors; release them first so
+        # empty_cache can return storage before the next token shape.
+        launch_raw = launch_triton = None
+        raw_tensors = triton_tensors = None
         torch.cuda.empty_cache()
 
     report = {
diff --git a/tools/ci/check.sh b/tools/ci/check.sh
deleted file mode 100755
index 316b5d5..0000000
--- a/tools/ci/check.sh
+++ /dev/null
@@ -1,21 +0,0 @@
-#!/usr/bin/env bash
-set -euo pipefail
-
-script_dir=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)
-repo_root=$(cd -- "$script_dir/../.." && pwd)
-python_bin=${PYTHON:-python3}
-
-export PYTHONPATH="$repo_root/compiler${PYTHONPATH:+:$PYTHONPATH}"
-cd "$repo_root"
-
-"$python_bin" -m compileall -q compiler/netra_compiler tools/compiler tools/ci tests/compiler
-"$python_bin" -m unittest discover -s tests/compiler -v
-"$python_bin" tools/ci/check_repository.py
-"$python_bin" tools/compiler/validate_gfx950_tactic_catalog.py
-"$python_bin" tools/ci/compile_examples.py
-
-while IFS= read -r -d '' shell_file; do
-  bash -n "$shell_file"
-done < <(git ls-files -z '*.sh')
-
-git diff --check
diff --git a/tools/ci/smoke_package.py b/tools/ci/smoke_package.py
index 0bed006..43944a6 100644
--- a/tools/ci/smoke_package.py
+++ b/tools/ci/smoke_package.py
@@ -6,7 +6,6 @@
 import argparse
 import os
 import subprocess
-import sys
 import tempfile
 import venv
 from pathlib import Path
diff --git a/tools/compiler/netra_compile.py b/tools/compiler/netra_compile.py
deleted file mode 100755
index 11746d3..0000000
--- a/tools/compiler/netra_compile.py
+++ /dev/null
@@ -1,13 +0,0 @@
-#!/usr/bin/env python3
-from __future__ import annotations
-
-import sys
-from pathlib import Path
-
-ROOT = Path(__file__).resolve().parents[2]
-sys.path.insert(0, str(ROOT / "compiler"))
-
-from netra_compiler.cli import main
-
-if __name__ == "__main__":
-    raise SystemExit(main())
diff --git a/tools/compiler/validate_engine.py b/tools/compiler/validate_engine.py
deleted file mode 100755
index 68da3e6..0000000
--- a/tools/compiler/validate_engine.py
+++ /dev/null
@@ -1,11 +0,0 @@
-#!/usr/bin/env python3
-from __future__ import annotations
-
-import sys
-from pathlib import Path
-
-ROOT = Path(__file__).resolve().parents[2]
-sys.path.insert(0, str(ROOT / "compiler"))
-from netra_compiler.cli import main
-
-if __name__ == "__main__": raise SystemExit(main(["validate", "--engine", *sys.argv[1:], "--static"]))
diff --git a/tools/profiling/rocprof_attach.py b/tools/profiling/rocprof_attach.py
index 8efa56d..719d2e2 100755
--- a/tools/profiling/rocprof_attach.py
+++ b/tools/profiling/rocprof_attach.py
@@ -10,7 +10,6 @@
 import ctypes
 import os
 import signal
-import sys
 import time
 
 DEFAULT_LIBRARY = (