Merge pull request #118 from codeflash-ai/azure_openai

CF-66 Move to Azure OpenAI service fer real tho

codeflash/main.py

@@ -20,7 +20,6 @@ from codeflash.code_utils.config_consts import (
 )
 from codeflash.code_utils.git_utils import get_repo_owner_and_name, get_github_secrets_page_url
 from codeflash.github.PrComment import FileDiffContent, PrComment
-from codeflash.verification import EXPLAIN_MODEL
 
 
 import os
@@ -225,10 +224,7 @@ class Optimizer:
                         code_to_optimize_with_dependents,
                         dependent_functions,
                     ) = get_constrained_function_context_and_dependent_functions(
-                        function_to_optimize,
-                        self.args.root,
-                        code_to_optimize,
-                        max_tokens=EXPLAIN_MODEL.max_tokens,
+                        function_to_optimize, self.args.root, code_to_optimize
                     )
                     logging.info("CODE TO OPTIMIZE %s", code_to_optimize_with_dependents)
                     module_path = module_name_from_file_path(path, self.args.root)

codeflash/optimization/function_context.py

@@ -168,11 +168,14 @@ def get_function_variables_definitions(
     return deduped_sources
 
 
+MAX_PROMPT_TOKENS = 4096  # 128000  # gpt-4-128k
+
+
 def get_constrained_function_context_and_dependent_functions(
     function_to_optimize: FunctionToOptimize,
     project_root_path: str,
     code_to_optimize: str,
-    max_tokens: int,
+    max_tokens: int = MAX_PROMPT_TOKENS,
 ) -> tuple[str, list[Source]]:
     # TODO: Not just do static analysis, but also find the datatypes of function arguments by running the existing
     #  unittests and inspecting the arguments to resolve the real definitions and dependencies.

codeflash/verification/__init__.py

@@ -1,42 +0,0 @@
-from pydantic.dataclasses import dataclass
-
-
-@dataclass
-class LLM:
-    name: str
-    max_tokens: int
-
-
-@dataclass
-class GPT_4_128k(LLM):
-    name: str = "gpt-4-1106-preview"
-    max_tokens: int = 128000
-
-
-@dataclass
-class GPT_4_32k(LLM):
-    name: str = "gpt4-32k"
-    max_tokens: int = 32768
-
-
-@dataclass
-class GPT_4(LLM):
-    name: str = "gpt-4"
-    max_tokens: int = 8192
-
-
-@dataclass
-class GPT_3_5_Turbo_16k(LLM):
-    name: str = "gpt-3.5-turbo-16k"
-    max_tokens: int = 16384
-
-
-@dataclass
-class GPT_3_5_Turbo(LLM):
-    name: str = "gpt-3.5-turbo"
-    max_tokens: int = 4096
-
-
-EXPLAIN_MODEL: LLM = GPT_4_128k()
-PLAN_MODEL = GPT_4_128k()
-EXECUTE_MODEL = GPT_4_128k()