aient 1.1.85__tar.gz → 1.1.87__tar.gz

{aient-1.1.85 → aient-1.1.87}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aient
-Version: 1.1.85
+Version: 1.1.87
 Summary: Aient: The Awakening of Agent.
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown

{aient-1.1.85 → aient-1.1.87}/aient/core/response.py

@@ -183,7 +183,7 @@ async def fetch_vertex_claude_response_stream(client, url, headers, payload, mod
             function_call_id = function_call["id"]
             sse_string = await generate_sse_response(timestamp, model, content=None, tools_id=function_call_id, function_call_name=function_call_name)
             yield sse_string
-            function_full_response = json.dumps(function_call["input"])
+            function_full_response = await asyncio.to_thread(json.dumps, function_call["input"])
             sse_string = await generate_sse_response(timestamp, model, content=None, tools_id=function_call_id, function_call_name=None, function_call_content=function_full_response)
             yield sse_string
 
@@ -213,6 +213,9 @@ async def fetch_gpt_response_stream(client, url, headers, payload, timeout):
             while "\n" in buffer:
                 line, buffer = buffer.split("\n", 1)
                 # logger.info("line: %s", repr(line))
+                if line.startswith(": keepalive"):
+                    yield line + end_of_line
+                    continue
                 if line and not line.startswith(":") and (result:=line.lstrip("data: ").strip()):
                     if result.strip() == "[DONE]":
                         break
@@ -306,7 +309,8 @@ async def fetch_gpt_response_stream(client, url, headers, payload, timeout):
                     else:
                         if no_stream_content:
                             del line["choices"][0]["message"]
-                        yield "data: " + json.dumps(line).strip() + end_of_line
+                        json_line = await asyncio.to_thread(json.dumps, line)
+                        yield "data: " + json_line.strip() + end_of_line
     yield "data: [DONE]" + end_of_line
 
 async def fetch_azure_response_stream(client, url, headers, payload, timeout):
@@ -363,7 +367,8 @@ async def fetch_azure_response_stream(client, url, headers, payload, timeout):
                     else:
                         if no_stream_content:
                             del line["choices"][0]["message"]
-                        yield "data: " + json.dumps(line).strip() + end_of_line
+                        json_line = await asyncio.to_thread(json.dumps, line)
+                        yield "data: " + json_line.strip() + end_of_line
     yield "data: [DONE]" + end_of_line
 
 async def fetch_cloudflare_response_stream(client, url, headers, payload, model, timeout):

{aient-1.1.85 → aient-1.1.87}/aient/core/utils.py

@@ -243,7 +243,8 @@ def parse_rate_limit(limit_string):
         'h': 3600, 'hr': 3600, 'hour': 3600,
         'd': 86400, 'day': 86400,
         'mo': 2592000, 'month': 2592000,
-        'y': 31536000, 'year': 31536000
+        'y': 31536000, 'year': 31536000,
+        'tpr': -1,
     }
 
     # 处理多个限制条件
@@ -377,6 +378,32 @@ class ThreadSafeCircularList:
                     logger.warning(f"All API keys are rate limited!")
                     raise HTTPException(status_code=429, detail="Too many requests")
 
+    async def is_tpr_exceeded(self, model: str = None, tokens: int = 0) -> bool:
+        """Checks if the request exceeds the TPR (Tokens Per Request) limit."""
+        if not tokens:
+            return False
+
+        async with self.lock:
+            rate_limit = None
+            model_key = model or "default"
+            if model and model_key in self.rate_limits:
+                rate_limit = self.rate_limits[model_key]
+            else:
+                # fuzzy match
+                for limit_model in self.rate_limits:
+                    if limit_model != "default" and model and limit_model in model:
+                        rate_limit = self.rate_limits[limit_model]
+                        break
+            if rate_limit is None:
+                rate_limit = self.rate_limits.get("default", [])
+
+            for limit_count, limit_period in rate_limit:
+                if limit_period == -1:  # TPR limit
+                    if tokens > limit_count:
+                        # logger.warning(f"API provider for model {model_key} exceeds TPR limit ({tokens}/{limit_count}).")
+                        return True
+        return False
+
     async def is_all_rate_limited(self, model: str = None) -> bool:
         """检查是否所有的items都被速率限制
 

{aient-1.1.85 → aient-1.1.87}/aient/models/chatgpt.py

@@ -37,10 +37,18 @@ class RateLimitError(Exception):
     """Custom exception for rate limit (429) errors."""
     pass
 
+class BadRequestError(Exception):
+    """Custom exception for bad request (400) errors."""
+    pass
+
 class HTTPError(Exception):
     """Custom exception for HTTP 500 errors."""
     pass
 
+class InputTokenCountExceededError(Exception):
+    """Custom exception for input token count exceeding the maximum."""
+    pass
+
 class ConfigurationError(Exception):
     """Custom exception for configuration errors."""
     pass
@@ -802,6 +810,10 @@ class chatgpt(BaseLLM):
                             raise ModelNotFoundError(f"Model: {model or self.engine} not found!")
                         if "HTTP Error', 'status_code': 429" in processed_chunk:
                             raise RateLimitError(f"Rate limit exceeded for model: {model or self.engine}")
+                        if "HTTP Error', 'status_code': 413" in processed_chunk:
+                            raise InputTokenCountExceededError(processed_chunk)
+                        if "HTTP Error', 'status_code': 400" in processed_chunk:
+                            raise BadRequestError(f"Bad Request: {processed_chunk}")
                         if "HTTP Error', 'status_code': " in processed_chunk:
                             raise HTTPError(f"HTTP Error: {processed_chunk}")
                     yield processed_chunk
@@ -826,6 +838,12 @@ class chatgpt(BaseLLM):
             except RateLimitError as e:
                 self.logger.warning(f"{e}, retrying...")
                 continue
+            except InputTokenCountExceededError as e:
+                self.logger.error(f"The request body is too long: {e}")
+                raise
+            except BadRequestError as e:
+                self.logger.error(f"Bad request error: {e}")
+                raise
             except ValidationError as e:
                 self.logger.warning(f"Validation failed: {e}. Retrying with corrective prompt.")
                 need_done_prompt = [

{aient-1.1.85 → aient-1.1.87}/aient.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: aient
-Version: 1.1.85
+Version: 1.1.87
 Summary: Aient: The Awakening of Agent.
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown

{aient-1.1.85 → aient-1.1.87}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "aient"
-version = "1.1.85"
+version = "1.1.87"
 description = "Aient: The Awakening of Agent."
 readme = "README.md"
 requires-python = ">=3.11"