Merge pull request #1 from TonicAI/f/squash-test

refactor

src/runner.py

@@ -3,57 +3,49 @@
 from typing import List
 from tonic_validate import BenchmarkItem, LLMResponse, ValidateApi, ValidateScorer
 
-path_to_responses = os.environ.get('VALIDATE_RESPONSES_PATH', None)
-validate_api_key = os.environ.get('TONIC_VALIDATE_SERVER_API_KEY', None)
-validate_project_id = os.environ.get('TONIC_VALIDATE_SERVER_PROJECT_ID', None)
-
-if validate_api_key is None:
-    exit('Error: You must specify TONIC_VALIDATE_SERVER_API_KEY, the API key for the Tonic Validate server')
-
-if validate_project_id is None:
-    exit('Error: You must specify TONIC_VALIDATE_SERVER_PROJECT_ID, the project ID for the Tonic Validate server')
-
-if path_to_responses is None:
-    exit('Error: You must specify VALIDATE_RESPONSES_PATH, the path to your LLM question and responses')
+def get_env_variable(var_name: str, error_message: str) -> str:
+    """Retrieve an environment variable or exit with an error message."""
+    value = os.environ.get(var_name)
+    if not value:
+        exit(error_message)
+    return value
+
+def load_responses(path: str) -> List[LLMResponse]:
+    """Load LLM responses from a JSON file."""
+    with open(path, 'r') as json_data:
+        try:
+            responses = json.load(json_data)
+        except ValueError as e:
+            exit(f'Error: Failed to parse {path}, please ensure it is valid JSON. Error: {e}')
+    return [
+        LLMResponse(
+            response['llm_answer'],
+            response.get('llm_context_list', []),
+            benchmark_item=BenchmarkItem(response['benchmark_item']['question'], response['benchmark_item']['answer'])
+        ) for response in responses
+    ]
+
+validate_api_key = get_env_variable('TONIC_VALIDATE_SERVER_API_KEY', 'Error: TONIC_VALIDATE_SERVER_API_KEY is required.')
+validate_project_id = get_env_variable('TONIC_VALIDATE_SERVER_PROJECT_ID', 'Error: TONIC_VALIDATE_SERVER_PROJECT_ID is required.')
+path_to_responses = get_env_variable('VALIDATE_RESPONSES_PATH', 'Error: VALIDATE_RESPONSES_PATH is required.')
 
 if not os.path.exists(path_to_responses):
-    exit('ERROR: The VALIDATE_RESPONSES_PATH provided ("{}") does not exist'.format(path_to_responses))
-
-openai_key = os.environ.get("OPENAI_API_KEY")
-azure_key = os.environ.get("AZURE_OPENAI_KEY")
-azure_endpoint = os.environ.get("AZURE_OPENAI_ENDPOINT")
-
-if openai_key is None or openai_key=='':
-    if (azure_key is None or azure_key=='') and (azure_endpoint is None or azure_endpoint==''):
-        exit('ERROR: You must set either an OpenAI key or an Azure key and Azure endpoint')
-
+    exit(f'ERROR: The path {path_to_responses} does not exist.')
 
-with open(path_to_responses) as json_data:
-    try:
-        responses = json.load(json_data)
-        json_data.close()
-    except ValueError:
-        exit('Error: Failed to parse {}, please ensure it is valid JSON'.format(path_to_responses))
-
-llm_responses: List[LLMResponse] = []
-for response in responses:
-    llm_context_list = []
-    if 'llm_context_list' in response:
-        llm_context_list = response['llm_context_list']
-
-    l = LLMResponse(response['llm_answer'], llm_context_list, benchmark_item=BenchmarkItem(response['benchmark_item']['question'], response['benchmark_item']['answer']))
-    llm_responses.append(l)
+llm_responses = load_responses(path_to_responses)
 
 scorer = ValidateScorer()
 run = scorer.score_responses(llm_responses)
 
-github_server = os.environ.get('GITHUB_SERVER_URL')
-github_repository = os.environ.get('GITHUB_REPOSITORY')
-github_sha = os.environ.get('GITHUB_SHA')
+validate_api = ValidateApi(validate_api_key)
+
+# Use default github action environment variables
+github_server = os.environ.get('GITHUB_SERVER_URL', None)
+github_repository = os.environ.get('GITHUB_REPOSITORY', None)
+github_sha = os.environ.get('GITHUB_SHA', None)
 
 run_metadata = {
     'commit_url': f"{github_server}/{github_repository}/commit/{github_sha}"
 }
 
-validate_api = ValidateApi(validate_api_key)
 validate_api.upload_run(validate_project_id, run, run_metadata)