vectorvein 0.2.86__tar.gz → 0.2.87__tar.gz

{vectorvein-0.2.86 → vectorvein-0.2.87}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vectorvein
-Version: 0.2.86
+Version: 0.2.87
 Summary: VectorVein Python SDK
 Author-Email: Anderson <andersonby@163.com>
 License: MIT

{vectorvein-0.2.86 → vectorvein-0.2.87}/pyproject.toml

@@ -17,7 +17,7 @@ description = "VectorVein Python SDK"
 name = "vectorvein"
 readme = "README.md"
 requires-python = ">=3.10"
-version = "0.2.86"
+version = "0.2.87"
 
 [project.license]
 text = "MIT"

{vectorvein-0.2.86 → vectorvein-0.2.87}/src/vectorvein/chat_clients/utils.py

@@ -103,18 +103,8 @@ def convert_type(value, value_type):
 
 def get_token_counts(text: str | dict, model: str = "", use_token_server_first: bool = True) -> int:
     if use_token_server_first and settings.token_server is not None:
-        base_url = (
-            settings.token_server.url
-            if settings.token_server.url is not None
-            else f"http://{settings.token_server.host}:{settings.token_server.port}"
-        )
-        _, response = (
-            Retry(httpx.post)
-            .args(url=f"{base_url}/count_tokens", json={"text": text, "model": model}, timeout=None)
-            .retry_times(5)
-            .sleep_time(1)
-            .run()
-        )
+        base_url = settings.token_server.url if settings.token_server.url is not None else f"http://{settings.token_server.host}:{settings.token_server.port}"
+        _, response = Retry(httpx.post).args(url=f"{base_url}/count_tokens", json={"text": text, "model": model}, timeout=None).retry_times(5).sleep_time(1).run()
         if response is not None:
             try:
                 result = response.json()
@@ -147,13 +137,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
             ],
         }
 
-        _, response = (
-            Retry(httpx.post)
-            .args(url=tokenize_url, headers=headers, json=request_body, timeout=None)
-            .retry_times(5)
-            .sleep_time(10)
-            .run()
-        )
+        _, response = Retry(httpx.post).args(url=tokenize_url, headers=headers, json=request_body, timeout=None).retry_times(5).sleep_time(10).run()
         if response is None:
             return 1000
         result = response.json()
@@ -174,13 +158,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
                 {"role": "user", "content": text},
             ],
         }
-        _, response = (
-            Retry(httpx.post)
-            .args(url=tokenize_url, headers=headers, json=request_body, timeout=None)
-            .retry_times(5)
-            .sleep_time(10)
-            .run()
-        )
+        _, response = Retry(httpx.post).args(url=tokenize_url, headers=headers, json=request_body, timeout=None).retry_times(5).sleep_time(10).run()
         if response is None:
             return 1000
         result = response.json()
@@ -194,11 +172,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
             endpoint_id = endpoint_id["endpoint_id"]
         endpoint = settings.get_endpoint(endpoint_id)
 
-        api_base = (
-            endpoint.api_base.removesuffix("/openai/")
-            if endpoint.api_base
-            else "https://generativelanguage.googleapis.com/v1beta"
-        )
+        api_base = endpoint.api_base.removesuffix("/openai/") if endpoint.api_base else "https://generativelanguage.googleapis.com/v1beta"
         base_url = f"{api_base}/models/{backend_setting.id}:countTokens"
         params = {"key": endpoint.api_key}
         request_body = {
@@ -209,13 +183,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
                 ],
             },
         }
-        _, response = (
-            Retry(httpx.post)
-            .args(base_url, json=request_body, params=params, timeout=None)
-            .retry_times(5)
-            .sleep_time(10)
-            .run()
-        )
+        _, response = Retry(httpx.post).args(base_url, json=request_body, params=params, timeout=None).retry_times(5).sleep_time(10).run()
         if response is None:
             return 1000
         result = response.json()
@@ -230,12 +198,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
                     endpoint_id = endpoint_choice
                 endpoint = settings.get_endpoint(endpoint_id)
 
-                if (
-                    endpoint.is_vertex
-                    or endpoint.is_bedrock
-                    or endpoint.endpoint_type == "anthropic_vertex"
-                    or endpoint.endpoint_type == "anthropic_bedrock"
-                ):
+                if endpoint.is_vertex or endpoint.is_bedrock or endpoint.endpoint_type == "anthropic_vertex" or endpoint.endpoint_type == "anthropic_bedrock":
                     continue
                 elif endpoint.endpoint_type in ("default", "anthropic"):
                     return (
@@ -277,13 +240,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
                 {"role": "user", "content": text},
             ],
         }
-        _, response = (
-            Retry(httpx.post)
-            .args(url=tokenize_url, headers=headers, json=request_body, timeout=None)
-            .retry_times(5)
-            .sleep_time(10)
-            .run()
-        )
+        _, response = Retry(httpx.post).args(url=tokenize_url, headers=headers, json=request_body, timeout=None).retry_times(5).sleep_time(10).run()
         if response is None:
             return 1000
         result = response.json()
@@ -306,13 +263,7 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
                 {"role": "user", "content": text},
             ],
         }
-        _, response = (
-            Retry(httpx.post)
-            .args(url=tokenize_url, headers=headers, json=request_body, timeout=None)
-            .retry_times(5)
-            .sleep_time(10)
-            .run()
-        )
+        _, response = Retry(httpx.post).args(url=tokenize_url, headers=headers, json=request_body, timeout=None).retry_times(5).sleep_time(10).run()
         if response is None:
             return 1000
         result = response.json()
@@ -453,48 +404,52 @@ def cutoff_messages(
     return system_message + messages
 
 
-def format_image_message(image: str, backend: BackendType = BackendType.OpenAI) -> dict:
-    from ..utilities.media_processing import ImageProcessor
+def format_image_message(image: str, backend: BackendType = BackendType.OpenAI, process_image: bool = True) -> dict:
+    if process_image:
+        from ..utilities.media_processing import ImageProcessor
 
-    image_processor = ImageProcessor(image_source=image)
-    if backend == BackendType.OpenAI:
-        return {
-            "type": "image_url",
-            "image_url": {"url": image_processor.data_url},
-        }
-    elif backend == BackendType.Anthropic:
-        return {
-            "type": "image",
-            "source": {
-                "type": "base64",
-                "media_type": image_processor.mime_type,
-                "data": image_processor.base64_image,
-            },
-        }
+        image_processor = ImageProcessor(image_source=image)
+        if backend == BackendType.OpenAI:
+            return {
+                "type": "image_url",
+                "image_url": {"url": image_processor.data_url},
+            }
+        elif backend == BackendType.Anthropic:
+            return {
+                "type": "image",
+                "source": {
+                    "type": "base64",
+                    "media_type": image_processor.mime_type,
+                    "data": image_processor.base64_image,
+                },
+            }
+        else:
+            return {
+                "type": "image_url",
+                "image_url": {"url": image_processor.data_url},
+            }
     else:
-        return {
-            "type": "image_url",
-            "image_url": {"url": image_processor.data_url},
-        }
+        if backend == BackendType.Anthropic:
+            return {"type": "image", "source": {"type": "base64", "media_type": "image/jpeg", "data": image}}
+        else:
+            return {"type": "image_url", "image_url": {"url": image}}
 
 
 def format_workflow_messages(message: dict, content: str, backend: BackendType):
     formatted_messages = []
 
     # 工具调用消息
-    if backend in (BackendType.OpenAI, BackendType.ZhiPuAI, BackendType.Mistral, BackendType.Yi, BackendType.Gemini):
+    if backend in (BackendType.OpenAI, BackendType.ZhiPuAI, BackendType.Mistral, BackendType.Yi, BackendType.Gemini, BackendType.DeepSeek):
         tool_call_message = {
-            "content": None,
+            "content": content,
             "role": "assistant",
             "tool_calls": [
                 {
-                    "id": message["metadata"]["selected_workflow"]["tool_call_id"],
+                    "id": message["metadata"]["selected_workflow"]["tool_call_id"] or message["metadata"]["record_id"],
                     "type": "function",
                     "function": {
                         "name": message["metadata"]["selected_workflow"]["function_name"],
-                        "arguments": json.dumps(
-                            message["metadata"]["selected_workflow"]["params"], ensure_ascii=False
-                        ),
+                        "arguments": json.dumps(message["metadata"]["selected_workflow"]["params"], ensure_ascii=False),
                     },
                 }
             ],
@@ -527,10 +482,10 @@ def format_workflow_messages(message: dict, content: str, backend: BackendType):
     formatted_messages.append(tool_call_message)
 
     # 工具调用结果消息
-    if backend in (BackendType.OpenAI, BackendType.ZhiPuAI, BackendType.Mistral, BackendType.Yi, BackendType.Gemini):
+    if backend in (BackendType.OpenAI, BackendType.ZhiPuAI, BackendType.Mistral, BackendType.Yi, BackendType.Gemini, BackendType.DeepSeek):
         tool_call_result_message = {
             "role": "tool",
-            "tool_call_id": message["metadata"]["selected_workflow"]["tool_call_id"],
+            "tool_call_id": message["metadata"]["selected_workflow"]["tool_call_id"] or message["metadata"]["record_id"],
             "name": message["metadata"]["selected_workflow"]["function_name"],
             "content": message["metadata"].get("workflow_result", ""),
         }
@@ -558,8 +513,8 @@ def format_workflow_messages(message: dict, content: str, backend: BackendType):
         }
     formatted_messages.append(tool_call_result_message)
 
-    if content and backend not in (BackendType.Mistral, BackendType.Anthropic):
-        formatted_messages.append({"role": "assistant", "content": content})
+    # if content and backend not in (BackendType.Mistral, BackendType.Anthropic):
+    #     formatted_messages.append({"role": "assistant", "content": content})
 
     return formatted_messages
 
@@ -569,6 +524,7 @@ def transform_from_openai_message(
     backend: BackendType,
     native_multimodal: bool = False,
     function_call_available: bool = False,
+    process_image: bool = True,
 ):
     role = message.get("role", "user")
     content = message.get("content", "")
@@ -593,7 +549,7 @@ def transform_from_openai_message(
                     formatted_content.append({"type": "text", "text": item})
                 elif isinstance(item, dict) and "type" in item:
                     if item["type"] == "image_url":
-                        formatted_content.append(format_image_message(item["image_url"]["url"], backend))
+                        formatted_content.append(format_image_message(item["image_url"]["url"], backend, process_image))
                     else:
                         formatted_content.append(item)
             if tool_calls:
@@ -674,6 +630,7 @@ def format_messages(
     backend: BackendType = BackendType.OpenAI,
     native_multimodal: bool = False,
     function_call_available: bool = False,
+    process_image: bool = True,
 ) -> list:
     """将 VectorVein 和 OpenAI 的 Message 序列化后的格式转换为不同模型支持的格式
 
@@ -681,6 +638,8 @@ def format_messages(
         messages (list): VectorVein Or OpenAI messages list.
         backend (str, optional): Messages format target backend. Defaults to BackendType.OpenAI.
         native_multimodal (bool, optional): Use native multimodal ability. Defaults to False.
+        function_call_available (bool, optional): Use function call ability. Defaults to False.
+        process_image (bool, optional): Process image. Defaults to True.
 
     Returns:
         list: 转换后的消息列表
@@ -697,17 +656,13 @@ def format_messages(
             content = message["content"]["text"]
             if message["content_type"] == "TXT":
                 role = "user" if message["author_type"] == "U" else "assistant"
-                formatted_message = format_text_message(
-                    content, role, message.get("attachments", []), backend, native_multimodal
-                )
+                formatted_message = format_text_message(content, role, message.get("attachments", []), backend, native_multimodal, process_image)
                 formatted_messages.append(formatted_message)
             elif message["content_type"] == "WKF" and message["status"] in ("S", "R"):
                 formatted_messages.extend(format_workflow_messages(message, content, backend))
         else:
             # 处理 OpenAI 格式的消息
-            formatted_message = transform_from_openai_message(
-                message, backend, native_multimodal, function_call_available
-            )
+            formatted_message = transform_from_openai_message(message, backend, native_multimodal, function_call_available, process_image)
             formatted_messages.append(formatted_message)
 
     return formatted_messages
@@ -719,6 +674,7 @@ def format_text_message(
     attachments: list,
     backend: BackendType,
     native_multimodal: bool,
+    process_image: bool = True,
 ):
     images_extensions = ("jpg", "jpeg", "png", "bmp")
     has_images = any(attachment.lower().endswith(images_extensions) for attachment in attachments)
@@ -733,7 +689,7 @@ def format_text_message(
             "content": [
                 {"type": "text", "text": content},
                 *[
-                    format_image_message(image=attachment, backend=backend)
+                    format_image_message(image=attachment, backend=backend, process_image=process_image)
                     for attachment in attachments
                     if attachment.lower().endswith(images_extensions)
                 ],