refactor(llm): async completion -> async acompletion

examples/invoice_organizer/invoice_organizer.py

@@ -127,7 +127,7 @@ async def _identify_invoice_message(self, ctx: Context, message: Message):
             ),
         ]
 
-        response = await ctx.llm.completion(
+        response = await ctx.llm.acompletion(
             messages=prompts,
             tools=self._processor.tools,
             tool_choice="auto",

examples/invoice_organizer/search_query_configs.py

@@ -146,7 +146,7 @@ async def save_configs(self, ctx: Context, **composed_configs: Dict[str, Any]):
             ),
         ]
 
-        response = await ctx.llm.completion(
+        response = await ctx.llm.acompletion(
             messages=messages,
             tools=[fn_reg.get_tool_param()],
             tool_choice="required",

npiai/context/configurator.py

@@ -123,7 +123,7 @@ async def _parse_instruction(self, ctx: Context, instruction: str):
             ),
         ]
 
-        response = await ctx.llm.completion(
+        response = await ctx.llm.acompletion(
             messages=messages,
             tools=[fn_reg.get_tool_param()],
             tool_choice="required",
@@ -174,7 +174,7 @@ async def _finalize_configs(self, ctx: Context, configs: Dict[str, Any]):
             ),
         ]
 
-        response = await ctx.llm.completion(
+        response = await ctx.llm.acompletion(
             messages=messages,
             tools=[fn_reg.get_tool_param()],
             tool_choice="required",

npiai/core/browser/_navigator.py

@@ -263,7 +263,7 @@ async def chat(self, ctx: Context, instruction: str) -> str:
                 return f"Maximum number of steps reached. Last response was: {response_str}"
 
     async def _call_llm(self, ctx: Context, task: Task) -> str:
-        response = await ctx.llm.completion(
+        response = await ctx.llm.acompletion(
             messages=task.conversations(),
             max_tokens=4096,
         )

npiai/core/tool/_agent.py

@@ -82,7 +82,7 @@ async def chat(
 
     async def _call_llm(self, ctx: Context, task: Task) -> str:
         while True:
-            response = await ctx.llm.completion(
+            response = await ctx.llm.acompletion(
                 messages=task.conversations(),
                 tools=self._tool.tools,
                 tool_choice="auto",

npiai/llm/llm.py

@@ -1,8 +1,7 @@
 from enum import Enum
-from typing import List, Union
 import os
 import asyncio
-from litellm import acompletion, ModelResponse, CustomStreamWrapper, drop_params
+from litellm import completion, acompletion, ModelResponse, CustomStreamWrapper
 
 
 class Provider(Enum):
@@ -27,13 +26,15 @@ def get_provider(self) -> Provider:
         return self.provider
 
     # TODO: kwargs typings
-    async def completion(self, **kwargs) -> Union[ModelResponse, CustomStreamWrapper]:
+    async def acompletion(self, **kwargs) -> ModelResponse | CustomStreamWrapper:
         return await acompletion(
             model=self.model, api_key=self.api_key, drop_params=True, **kwargs
         )
 
-    def completion_sync(self, **kwargs) -> Union[ModelResponse, CustomStreamWrapper]:
-        return asyncio.run(self.completion(**kwargs))
+    def completion(self, **kwargs) -> ModelResponse | CustomStreamWrapper:
+        return completion(
+            model=self.model, api_key=self.api_key, drop_params=True, **kwargs
+        )
 
 
 class OpenAI(LLM):

npiai/utils/llm_summarize.py

@@ -18,7 +18,7 @@ async def llm_summarize(
     final_response_content = ""
 
     while True:
-        response = await llm.completion(
+        response = await llm.acompletion(
             messages=messages_copy,
             max_tokens=4096,
             # use fixed temperature and seed to ensure deterministic results

npiai/utils/llm_tool_call.py

@@ -18,7 +18,7 @@ async def llm_tool_call(
     if fn_reg.model is None:
         raise RuntimeError("Unable to modeling tool function")
 
-    response = await llm.completion(
+    response = await llm.acompletion(
         messages=messages,
         tools=[fn_reg.get_tool_param()],
         max_tokens=4096,