diff --git a/pyproject.toml b/pyproject.toml
index 85a690c5..52b147ca 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -51,4 +51,13 @@ dev = [
 
 [tool.setuptools.packages.find]
 where = ["src"]
-include = ["kernelbench*"]
\ No newline at end of file
+include = ["kernelbench*"]
+
+[tool.setuptools]
+include-package-data = true
+
+[tool.setuptools.package-data]
+kernelbench = [
+  "prompts/*.toml",
+  "prompts/**/*.py",
+]
\ No newline at end of file
diff --git a/src/kernelbench/prompt_constructor_toml.py b/src/kernelbench/prompt_constructor_toml.py
index 7b8b1bdf..c278123e 100644
--- a/src/kernelbench/prompt_constructor_toml.py
+++ b/src/kernelbench/prompt_constructor_toml.py
@@ -1,5 +1,7 @@
 # src/prompt_constructor_toml.py | toml based prompt constructor
+import importlib.resources
 import os
+from pathlib import Path
 import runpy
 import tomli  
 from dataclasses import dataclass
@@ -15,10 +17,12 @@
 """
 
 REPO_TOP_PATH = os.path.abspath(os.path.join(os.path.dirname(__file__), "../.."))
-PROMPTS_TOML = os.path.join(REPO_TOP_PATH, "src/kernelbench/prompts/prompts.toml")
+PROMPTS_DIR = importlib.resources.files("kernelbench") / "prompts"
+assert PROMPTS_DIR.is_dir(), f"Prompts directory not found at {PROMPTS_DIR}"
 
-assert os.path.exists(PROMPTS_TOML), f"Prompts.toml not found at {PROMPTS_TOML}" 
-GPU_SPECS_PY = "src/kernelbench/prompts/hardware/gpu_specs.py"
+
+GPU_SPECS_PY = PROMPTS_DIR / "hardware" / "gpu_specs.py"
+assert GPU_SPECS_PY.is_file(), f"GPU specs file not found at {GPU_SPECS_PY}"
 HARDWARE_COMPONENT_KEYS = [
     "hardware_header",
     "hardware_specs",
@@ -50,7 +54,7 @@ class PromptConfig:
     data: Dict[str, Any]  # The raw parsed TOML data as nested dictionaries
 
     @classmethod
-    def from_toml(cls, path: str) -> "PromptConfig":
+    def from_toml(cls, path: str | Path) -> "PromptConfig":
         """
         Load and parse a TOML configuration file.
         
@@ -134,11 +138,11 @@ def _gpu_context_from_gpu_specs(py_path: str, gpu_name: str) -> Dict[str, str]:
 
 def render_prompt_by_option(
     *,
-    prompts_toml: str,
+    prompts_toml: Path,
     backend: str,
     option: str,
     context: Dict[str, str],
-    gpu_specs_py: Optional[str] = None,
+    gpu_specs_py: Optional[Path] = None,
     gpu_name: Optional[str] = None,
     precision: Optional[str] = None,
     include_hardware: bool = False,
@@ -292,11 +296,11 @@ def render_example_entry(input_code: str, output_code: str, example_label: str)
     
     # Load GPU details if requested
     if option_data.get("requires_gpu") or include_hardware:
-        if not (gpu_specs_py and gpu_name):
+        if not (gpu_specs_py and gpu_name and gpu_specs_py.is_file()):
             raise ValueError(
                 f"Hardware info requested for option '{option}'; provide gpu_specs_py and gpu_name"
             )
-        context = {**context, **_gpu_context_from_gpu_specs(_abs_path(gpu_specs_py), gpu_name)}
+        context = {**context, **_gpu_context_from_gpu_specs(gpu_specs_py, gpu_name)}
     
     # Builds the prompt from the components in the toml file. 
     prompt_parts = []
@@ -347,7 +351,7 @@ def get_prompt_for_backend(
         gpu_name: GPU identifier used when include_hardware is True (e.g., "A100")
     """
     return render_prompt_by_option(
-        prompts_toml=PROMPTS_TOML,
+        prompts_toml=PROMPTS_DIR / "prompts.toml",
         backend=backend.lower(),
         option=option.lower(),
         context={"ref_arch_src": ref_arch_src},
@@ -367,7 +371,7 @@ def get_custom_prompt(
     precision: Optional[str] = None,
     include_hardware: bool = False,
     gpu_name: Optional[str] = None,
-    prompts_toml: str = PROMPTS_TOML,
+    prompts_toml: Path = PROMPTS_DIR / "prompts.toml",
 ) -> str:
     """
     Render a prompt defined under [custom_prompts.<custom_key>] in prompts.toml.