vectorvein 0.1.29__tar.gz → 0.1.30__tar.gz

{vectorvein-0.1.29 → vectorvein-0.1.30}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vectorvein
-Version: 0.1.29
+Version: 0.1.30
 Summary: Default template for PDM package
 Author-Email: Anderson <andersonby@163.com>
 License: MIT

{vectorvein-0.1.29 → vectorvein-0.1.30}/pyproject.toml

@@ -16,7 +16,7 @@ description = "Default template for PDM package"
 name = "vectorvein"
 readme = "README.md"
 requires-python = ">=3.10"
-version = "0.1.29"
+version = "0.1.30"
 
 [project.license]
 text = "MIT"

{vectorvein-0.1.29 → vectorvein-0.1.30}/src/vectorvein/chat_clients/utils.py

@@ -21,8 +21,22 @@ from ..types.llm_parameters import (
 )
 
 
-chatgpt_encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")
-gpt_4o_encoding = tiktoken.encoding_for_model("gpt-4o")
+gpt_35_encoding = None
+gpt_4o_encoding = None
+
+
+def get_gpt_35_encoding():
+    global gpt_35_encoding
+    if gpt_35_encoding is None:
+        gpt_35_encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")
+    return gpt_35_encoding
+
+
+def get_gpt_4o_encoding():
+    global gpt_4o_encoding
+    if gpt_4o_encoding is None:
+        gpt_4o_encoding = tiktoken.encoding_for_model("gpt-4o")
+    return gpt_4o_encoding
 
 
 class ToolCallContentProcessor:
@@ -106,9 +120,9 @@ def get_token_counts(text: str | dict, model: str = "") -> int:
     if not isinstance(text, str):
         text = str(text)
     if model == "gpt-3.5-turbo":
-        return len(chatgpt_encoding.encode(text))
+        return len(get_gpt_35_encoding().encode(text))
     elif model in ("gpt-4o", "gpt-4o-mini"):
-        return len(gpt_4o_encoding.encode(text))
+        return len(get_gpt_4o_encoding().encode(text))
     elif model.startswith("abab"):
         model_setting = settings.minimax.models[model]
         if len(model_setting.endpoints) == 0:
@@ -140,7 +154,7 @@ def get_token_counts(text: str | dict, model: str = "") -> int:
     elif model in ("moonshot-v1-8k", "moonshot-v1-32k", "moonshot-v1-128k"):
         model_setting = settings.moonshot.models[model]
         if len(model_setting.endpoints) == 0:
-            return len(chatgpt_encoding.encode(text))
+            return len(get_gpt_35_encoding().encode(text))
         endpoint_id = model_setting.endpoints[0]
         endpoint = settings.get_endpoint(endpoint_id)
         tokenize_url = "https://api.moonshot.cn/v1/tokenizers/estimate-token-count"
@@ -165,7 +179,7 @@ def get_token_counts(text: str | dict, model: str = "") -> int:
     elif model.startswith("gemini"):
         model_setting = settings.gemini.models[model]
         if len(model_setting.endpoints) == 0:
-            return len(chatgpt_encoding.encode(text))
+            return len(get_gpt_35_encoding().encode(text))
         endpoint_id = model_setting.endpoints[0]
         endpoint = settings.get_endpoint(endpoint_id)
         url = f"{endpoint.api_base}/models/{model_setting.id}:countTokens"
@@ -198,7 +212,7 @@ def get_token_counts(text: str | dict, model: str = "") -> int:
     elif model.startswith("stepfun"):
         model_setting = settings.moonshot.models[model]
         if len(model_setting.endpoints) == 0:
-            return len(chatgpt_encoding.encode(text))
+            return len(get_gpt_35_encoding().encode(text))
         endpoint_id = model_setting.endpoints[0]
         endpoint = settings.get_endpoint(endpoint_id)
         tokenize_url = "https://api.stepfun.com/v1/token/count"
@@ -221,7 +235,7 @@ def get_token_counts(text: str | dict, model: str = "") -> int:
         result = response.json()
         return result["data"]["total_tokens"]
     else:
-        return len(chatgpt_encoding.encode(text))
+        return len(get_gpt_35_encoding().encode(text))
 
 
 def calculate_image_tokens(width: int, height: int, model: str = "gpt-4o"):