fix llm client by force

CTY-git · CTY-git · commit ce362bf1b8cb · 2025-02-14T00:13:31.000+08:00
diff --git a/patchwork/common/client/llm/aio.py b/patchwork/common/client/llm/aio.py
@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import os
+from pathlib import Path
 
 from openai.types.chat import (
     ChatCompletion,
@@ -54,10 +55,11 @@ def is_prompt_supported(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> int:
         for client in self.__clients:
             if client.is_model_supported(model):
-                return client.is_prompt_supported(
+                inputs = dict(
                     messages=messages,
                     model=model,
                     frequency_penalty=frequency_penalty,
@@ -74,6 +76,9 @@ def is_prompt_supported(
                     top_logprobs=top_logprobs,
                     top_p=top_p,
                 )
+                if file is not NotGiven:
+                    inputs["file"] = file
+                return client.is_prompt_supported(**inputs)
         return -1
 
     def truncate_messages(
@@ -101,27 +106,31 @@ def chat_completion(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> ChatCompletion:
         for client in self.__clients:
             if client.is_model_supported(model):
                 logger.debug(f"Using {client.__class__.__name__} for model {model}")
-                return client.chat_completion(
-                    messages,
-                    model,
-                    frequency_penalty,
-                    logit_bias,
-                    logprobs,
-                    max_tokens,
-                    n,
-                    presence_penalty,
-                    response_format,
-                    stop,
-                    temperature,
-                    tools,
-                    tool_choice,
-                    top_logprobs,
-                    top_p,
+                inputs = dict(
+                    messages=messages,
+                    model=model,
+                    frequency_penalty=frequency_penalty,
+                    logit_bias=logit_bias,
+                    logprobs=logprobs,
+                    max_tokens=max_tokens,
+                    n=n,
+                    presence_penalty=presence_penalty,
+                    response_format=response_format,
+                    stop=stop,
+                    temperature=temperature,
+                    tools=tools,
+                    tool_choice=tool_choice,
+                    top_logprobs=top_logprobs,
+                    top_p=top_p,
                 )
+                if file is not NotGiven:
+                    inputs["file"] = file
+                return client.chat_completion(**inputs)
         client_names = [client.__class__.__name__ for client in self.__original_clients]
         raise ValueError(
             f"Model {model} is not supported by {client_names} clients. "
diff --git a/patchwork/common/client/llm/anthropic.py b/patchwork/common/client/llm/anthropic.py
@@ -3,6 +3,7 @@
 import json
 import time
 from functools import lru_cache
+from pathlib import Path
 
 from anthropic import Anthropic
 from anthropic.types import Message, MessageParam, TextBlockParam
@@ -224,6 +225,7 @@ def is_prompt_supported(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> int:
         model_limit = self.__get_model_limit(model)
         input_kwargs = self.__adapt_chat_completion_request(
@@ -273,6 +275,7 @@ def chat_completion(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> ChatCompletion:
         input_kwargs = self.__adapt_chat_completion_request(
             messages=messages,
diff --git a/patchwork/common/client/llm/google.py b/patchwork/common/client/llm/google.py
@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import hashlib
 import time
 from functools import lru_cache
 from pathlib import Path
@@ -62,15 +63,16 @@ def __upload(self, file: Path | NotGiven) -> File | None:
         if file is NotGiven:
             return None
 
+        md5_name = hashlib.md5(file.read_bytes()).hexdigest()
         try:
-            file_ref = self.client.files.get(file.name)
+            file_ref = self.client.files.get(name=md5_name)
             if file_ref.error is None:
                 return file_ref
         except Exception as e:
             pass
 
         try:
-            file_ref = self.client.files.upload(file=file)
+            file_ref = self.client.files.upload(file=file, config=dict(name=md5_name))
             if file_ref.error is None:
                 return file_ref
         except Exception as e:
@@ -97,16 +99,16 @@ def is_prompt_supported(
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
         file: Path | NotGiven = NOT_GIVEN,
     ) -> int:
-        system, chat = self.__openai_messages_to_google_messages(messages)
+        system, contents = self.__openai_messages_to_google_messages(messages)
 
         file_ref = self.__upload(file)
         if file_ref is not None:
-            chat.append(file_ref)
+            contents.append(file_ref)
 
         try:
             token_response = self.client.models.count_tokens(
                 model=model,
-                contents=chat,
+                contents=contents,
                 config=CountTokensConfig(
                     system_instruction=system,
                 ),
@@ -155,6 +157,7 @@ def chat_completion(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> ChatCompletion:
         generation_dict = dict(
             stop_sequences=[stop] if isinstance(stop, str) else stop,
@@ -174,6 +177,9 @@ def chat_completion(
             )
 
         system_content, contents = self.__openai_messages_to_google_messages(messages)
+        file_ref = self.__upload(file)
+        if file_ref is not None:
+            contents.append(file_ref)
 
         response = self.client.models.generate_content(
             model=model,
diff --git a/patchwork/common/client/llm/openai_.py b/patchwork/common/client/llm/openai_.py
@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import functools
+from pathlib import Path
 
 import tiktoken
 from openai import OpenAI
@@ -82,6 +83,7 @@ def is_prompt_supported(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> int:
         # might not implement model endpoint
         if self.__is_not_openai_url():
@@ -125,6 +127,7 @@ def chat_completion(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> ChatCompletion:
         input_kwargs = dict(
             messages=messages,
diff --git a/patchwork/common/client/llm/protocol.py b/patchwork/common/client/llm/protocol.py
@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+from pathlib import Path
+
 from openai.types.chat import (
     ChatCompletion,
     ChatCompletionMessageParam,
@@ -51,6 +53,7 @@ def is_prompt_supported(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> int:
         ...
 
@@ -135,5 +138,6 @@ def chat_completion(
         tool_choice: ChatCompletionToolChoiceOptionParam | NotGiven = NOT_GIVEN,
         top_logprobs: Optional[int] | NotGiven = NOT_GIVEN,
         top_p: Optional[float] | NotGiven = NOT_GIVEN,
+        file: Path | NotGiven = NOT_GIVEN,
     ) -> ChatCompletion:
         ...
diff --git a/patchwork/steps/CallLLM/CallLLM.py b/patchwork/steps/CallLLM/CallLLM.py
@@ -121,14 +121,14 @@ def __call(self, prompts: list[list[dict]]) -> list[_InnerCallLLMResponse]:
             kwargs["file"] = Path(self.file)
 
         for prompt in prompts:
-            is_input_accepted = self.client.is_prompt_supported(prompt, self.model) > 0
+            is_input_accepted = self.client.is_prompt_supported(model=self.model, messages=prompt, **kwargs) > 0
             if not is_input_accepted:
                 self.set_status(StepStatus.WARNING, "Input token limit exceeded.")
                 prompt = self.client.truncate_messages(prompt, self.model)
 
             logger.trace(f"Message sent: \n{escape(indent(pformat(prompt), '  '))}")
             try:
-                completion = self.client.chat_completion(model=self.model, messages=prompt, **parsed_model_args)
+                completion = self.client.chat_completion(model=self.model, messages=prompt, **kwargs)
             except Exception as e:
                 logger.error(e)
                 completion = None
diff --git a/pyproject.toml b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "patchwork-cli"
-version = "0.0.99.dev3"
+version = "0.0.99.dev4"
 description = ""
 authors = ["patched.codes"]
 license = "AGPL"
