diff --git a/merico/unit_tests/find_reference_tests.py b/merico/unit_tests/find_reference_tests.py
index 64f707e..7be74ab 100644
--- a/merico/unit_tests/find_reference_tests.py
+++ b/merico/unit_tests/find_reference_tests.py
@@ -1,15 +1,112 @@
+import json
+from pathlib import Path
 from typing import List
 
-from assistants.directory_structure.relevant_file_finder import RelevantFileFinder
-from prompts import FIND_REFERENCE_PROMPT
-from tools.file_util import verify_file_list
+from devchat.llm.openai import chat_completion_no_stream_return_json
+from llm_conf import (
+    CONTEXT_SIZE,
+    DEFAULT_CONTEXT_SIZE,
+    DEFAULT_ENCODING,
+    USE_USER_MODEL,
+    USER_LLM_MODEL,
+)
+from openai_util import create_chat_completion_content
+from tools.file_util import (
+    is_not_hidden,
+    is_source_code,
+    is_test_file,
+    verify_file_list,
+)
+from tools.git_util import git_file_of_interest_filter
+from tools.tiktoken_util import get_encoding
+from tools.time_util import print_exec_time
+
+MODEL = USER_LLM_MODEL if USE_USER_MODEL else "gpt-4-turbo-preview"  # "gpt-3.5-turbo"
+ENCODING = (
+    get_encoding(DEFAULT_ENCODING)  # Use default encoding as an approximation
+    if USE_USER_MODEL
+    else get_encoding("cl100k_base")
+)
+TOKEN_BUDGET = int(CONTEXT_SIZE.get(MODEL, DEFAULT_CONTEXT_SIZE) * 0.95)
 
 
-def find_reference_tests(root_path: str, function_name: str, file_path: str) -> List[str]:
+FIND_REF_TEST_PROMPT = """
+As an advanced AI coding assistant,
+you're given the task to identify suitable reference test files that can be used as a guide
+for writing test cases for a specific function in the codebase.
+
+You're provided with a list of test files in the repository.
+Infer the purpose of each test file and identify the top 3 key files
+that may be relevant to the target function and can serve as a reference for writing test cases.
+The reference could provide a clear example of best practices
+in testing functions of a similar nature.
+
+The target function is {function_name}, located in the file {file_path}.
+The list of test files in the repository is as follows:
+
+{test_files_str}
+
+
+Answer in JSON format with a list of the top 3 key file paths under the key `files`.
+Make sure each file path is from the list of test files provided above.
+
+Example:
+{{
+    "files": ["<file path 1>", "<file path 2>", "<file path 3>"]
+}}
+
+"""
+
+
+def get_test_files(repo_root: str) -> List[str]:
+    """
+    Get all test files in the repository.
+    """
+    root = Path(repo_root)
+    is_git_interest = git_file_of_interest_filter(repo_root)
+
+    files = []
+    for filepath in root.rglob("*"):
+        relpath = filepath.relative_to(root)
+
+        is_candidate = (
+            filepath.is_file()
+            and is_not_hidden(relpath)
+            and is_git_interest(relpath)
+            and is_source_code(str(filepath), only_code=True)
+            and is_test_file(str(relpath))
+        )
+
+        if not is_candidate:
+            continue
+
+        files.append(str(relpath))
+
+    return files
+
+
+def _mk_user_msg(function_name: str, file_path: str, test_files: List[str]) -> str:
+    """
+    Create a user message to be sent to the model within the token budget.
+    """
+    test_files_str = "\n".join([f"- {f}" for f in test_files])
+    msg = FIND_REF_TEST_PROMPT.format(
+        function_name=function_name,
+        file_path=file_path,
+        test_files_str=test_files_str,
+    )
+
+    # TODO: check if the message fits within the token budget
+    # and adjust the content accordingly
+    return msg
+
+
+@print_exec_time("Model response time")
+def find_reference_tests(repo_root: str, function_name: str, file_path: str) -> List[str]:
     """Find reference tests for a specified function
 
     Args:
-        root_path (str): The path to the root directory of the codebase.
+        repo_root (str): The path to the root directory of the codebase.
         function_name (str): The name of the function to generate test cases for.
         file_path (str): The path to the file containing the target function
                         for which test cases will be generated.
@@ -18,11 +115,39 @@ def find_reference_tests(root_path: str, function_name: str, file_path: str) ->
         List[str]: A list of paths to files that may contain a reference test
                     for the specified function.
     """
-    finder = RelevantFileFinder(root_path=root_path)
-    objective = FIND_REFERENCE_PROMPT.format(
+    test_files = get_test_files(repo_root)
+
+    user_msg = _mk_user_msg(
         function_name=function_name,
         file_path=file_path,
+        test_files=test_files,
     )
 
-    test_paths = finder.analyze(objective)
-    return verify_file_list(test_paths, root_path)
+    json_res = {}
+    if USE_USER_MODEL:
+        # Use the wrapped api parameters
+        json_res = (
+            chat_completion_no_stream_return_json(
+                messages=[{"role": "user", "content": user_msg}],
+                llm_config={
+                    "model": MODEL,
+                    "temperature": 0.1,
+                },
+            )
+            or {}
+        )
+
+    else:
+        # Use the openai api parameters
+        content = create_chat_completion_content(
+            model=MODEL,
+            messages=[{"role": "user", "content": user_msg}],
+            response_format={"type": "json_object"},
+            temperature=0.1,
+        )
+        json_res = json.loads(content)
+
+    files = json_res.get("files", [])
+    ref_files = verify_file_list(files, repo_root)
+
+    return ref_files