Move all configuration to ApiViewReviewer constructor. (Azure#10395)

Author: tjprescott

packages/python-packages/apiview-copilot/cli.py

@@ -55,12 +55,14 @@ def local_review(
     """
     from src._apiview_reviewer import ApiViewReview
 
-    rg = ApiViewReview(language=language, model=model)
+    rg = ApiViewReview(
+        language=language, model=model, chunk_input=chunk_input, use_rag=use_rag
+    )
     filename = os.path.splitext(os.path.basename(path))[0]
 
     with open(path, "r") as f:
         apiview = f.read()
-    review = rg.get_response(apiview, chunk_input=chunk_input, use_rag=use_rag)
+    review = rg.get_response(apiview)
     output_path = os.path.join("scratch", "output", language)
     os.makedirs(output_path, exist_ok=True)
     output_file = os.path.join(output_path, f"{filename}.json")

packages/python-packages/apiview-copilot/evals/run.py

@@ -15,7 +15,6 @@
 
 dotenv.load_dotenv()
 
-MODEL = "o3-mini"
 NUM_RUNS: int = 3
 
 
@@ -101,8 +100,8 @@ def review_apiview(query: str, language: str):
         ApiViewReview,
     )
 
-    ai_review = ApiViewReview(language=language, model=MODEL)
-    review = ai_review.get_response(query, chunk_input=False, use_rag=False)
+    ai_review = ApiViewReview(language=language)
+    review = ai_review.get_response(query)
     return {"response": review.model_dump_json()}
 
 

packages/python-packages/apiview-copilot/src/_apiview_reviewer.py

@@ -60,20 +60,30 @@
 supported_models = [x for x in model_map.keys()]
 
 DEFAULT_MODEL = "o3-mini"
+DEFAULT_USE_RAG = False
+DEFAULT_CHUNK_INPUT = False
 
 
 class ApiViewReview:
 
-    def __init__(self, *, language: str, model: str = DEFAULT_MODEL):
-        self.language = language
-        self.model = model
-        self.search = SearchManager(language=language)
-        self.output_parser = ReviewResult
-        self.semantic_search_failed = False
+    def __init__(
+        self,
+        *,
+        language: str,
+        model: str = DEFAULT_MODEL,
+        use_rag: bool = DEFAULT_USE_RAG,
+        chunk_input: bool = DEFAULT_CHUNK_INPUT,
+    ):
         if model not in supported_models:
             raise ValueError(
                 f"Model {model} not supported. Supported models are: {', '.join(supported_models)}"
             )
+        self.language = language
+        self.model = model
+        self.use_rag = use_rag
+        self.chunk_input = chunk_input
+        self.search = SearchManager(language=language)
+        self.semantic_search_failed = False
 
     def _hash(self, obj) -> str:
         return str(hash(json.dumps(obj)))
@@ -89,13 +99,11 @@ def _ensure_env_vars(self, vars: List[str]):
         if missing:
             raise ValueError(f"Environment variables not set: {', '.join(missing)}")
 
-    def get_response(
-        self, apiview: str, *, chunk_input: bool = False, use_rag: bool = False
-    ) -> ReviewResult:
+    def get_response(self, apiview: str) -> ReviewResult:
         print(f"Generating review...")
 
         logger.info(
-            f"Starting review with model: {self.model}, language: {self.language}, RAG: {use_rag}"
+            f"Starting review with model: {self.model}, language: {self.language}, RAG: {self.use_rag}"
         )
 
         start_time = time()
@@ -106,7 +114,9 @@ def get_response(
         static_guideline_ids = [x["id"] for x in static_guidelines]
 
         # Prepare the document
-        chunked_apiview = SectionedDocument(apiview.splitlines(), chunk=chunk_input)
+        chunked_apiview = SectionedDocument(
+            apiview.splitlines(), chunk=self.chunk_input
+        )
         final_results = ReviewResult(
             guideline_ids=static_guideline_ids, status="Success", violations=[]
         )
@@ -179,7 +189,7 @@ def process_chunk(chunk_info):
 
                     try:
                         # build the context string
-                        if use_rag:
+                        if self.use_rag:
                             context = self._retrieve_and_resolve_guidelines(str(chunk))
                             if context:
                                 context_string = context.to_markdown()