publish next version

Author: trisongz

CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelogs
 
+#### v0.0.53 (2024-05-31)
+
+- Added support for new `gpt-4-o` model
+- Added support for external providers
+- Update certain validation behavior of Functions
+
+- TODO: Add batch create support
+
 #### v0.0.52 (2024-02-28)
 
 - Added support for the following parameters in `model_configurations` in `OpenAIManager`:

async_openai/manager.py

@@ -765,9 +765,21 @@ def register_client_endpoints(self):  # sourcery skip: low-code-quality
                 config['api_base'] = self.settings.proxy.endpoint
                 _has_proxy = True
             c = self.init_api_client(name, is_azure = is_azure, set_as_default = is_default, **config)
-            msg = f'Registered: `|g|{c.name}|e|` @ `{source_endpoint or c.base_url}` (Azure: {c.is_azure}, Px: {_has_proxy}'
-            if has_weights: msg += f', Weight: {client_weight}'
-            msg += ')'
+            msg = f'Registered: `|g|{c.name}|e|` @ `{source_endpoint or c.base_url}`'
+            extra_msgs = []
+            if has_weights: 
+                if isinstance(client_weight, float):
+                    _wp, _wsfx = '|g|', '|e|'
+                    if client_weight <= 0.0: 
+                        _wp, _wsfx = '', ''
+                    elif client_weight <= 0.25: _wp = '|r|'
+                    elif client_weight <= 0.45: _wp = '|y|'
+                    extra_msgs.append(f'Weight: {_wp}{client_weight}{_wsfx}')
+                else:
+                    extra_msgs.append(f'Weight: {client_weight}')
+            if c.is_azure: extra_msgs.append('Azure')
+            if _has_proxy: extra_msgs.append('Proxied')
+            if extra_msgs: msg += f' ({", ".join(extra_msgs)})'
             logger.info(msg, colored = True)
 
         # Set the models for inclusion

async_openai/types/functions.py

@@ -524,7 +524,7 @@ def parse_response(
         """
         schema = schema or self.schema
         try:
-            result = schema.model_validate(response.function_results[0].arguments, from_attributes = True)
+            result = schema.model_validate(response.function_results[0].arguments, from_attributes = True, context = {'source': 'function'})
             result._set_values_from_response(response, name = self.name if include_name else None, client_name = client_name)
             return result
         except IndexError as e:
@@ -533,7 +533,7 @@ def parse_response(
         except Exception as e:
             self.autologger.error(f"[{self.name} - {response.model} - {response.usage}] Failed to parse object: {e}\n{response.text}\n{response.function_results[0].arguments}")
             try:
-                result = schema.model_validate(resolve_json(response.function_results[0].arguments), from_attributes = True)
+                result = schema.model_validate(resolve_json(response.function_results[0].arguments), from_attributes = True, context = {'source': 'function'})
                 result._set_values_from_response(response, name = self.name if include_name else None)
                 return result
             except Exception as e:

async_openai/types/options.py

@@ -194,6 +194,7 @@ class FilePurpose(str, Enum):
     fine_tune = "fine-tune"
     train = "fine-tune-train"
     search = "search"
+    batch = "batch"
 
     @classmethod
     def parse_str(cls, value: Union[str, 'FilePurpose'], raise_error: bool = True):
@@ -206,6 +207,8 @@ def parse_str(cls, value: Union[str, 'FilePurpose'], raise_error: bool = True):
             return cls.fine_tune
         elif "search" in value:
             return cls.search
+        elif "batch" in value:
+            return cls.batch
         if not raise_error: return None
         raise ValueError(f"Cannot convert {value} to FilePurpose")
 

async_openai/types/pricing.yaml

@@ -24,6 +24,32 @@ gpt-4-0125-preview:
   endpoints:
   - chat
 
+gpt-4-turbo-2024-04-09:
+  aliases:
+  - gpt-4-turbo-2024
+  - gpt-4-turbo-2024-04-09
+  - gpt-4-2024-preview
+  context_length: 128000
+  costs:
+    unit: 1_000_000
+    input: 10.00
+    output: 30.00
+  endpoints:
+  - chat
+
+gpt-4o:
+  aliases:
+  - gpt-4-o
+  - gpt4o
+  - gpt-4o-2024-05-13
+  context_length: 128000
+  costs:
+    unit: 1_000_000
+    input: 5.00
+    output: 15.00
+  endpoints:
+  - chat
+
 gpt-4:
   aliases:
   - gpt-4-0613

async_openai/types/resources.py

@@ -2,11 +2,13 @@
 import json
 import aiohttpx
 import datetime
+import tempfile
+import pathlib
 from pydantic import ConfigDict
 from pydantic.types import ByteSize
 from lazyops.types import BaseModel, validator, lazyproperty
 from lazyops.types.models import get_pyd_field_names, pyd_parse_obj, get_pyd_dict, _BaseModel
-from lazyops.utils import ObjectDecoder
+from lazyops.utils import ObjectDecoder, ObjectEncoder
 from async_openai.utils.logs import logger
 from async_openai.utils.helpers import aparse_stream, parse_stream
 
@@ -156,6 +158,29 @@ def create_resource(
         resource_obj = resource.parse_obj(resource_kwargs)
         return resource_obj, return_kwargs
 
+
+    @staticmethod
+    def create_batch_resource(
+        resource: Type['BaseResource'],
+        batch: List[Union[Dict[str, Any], Any]],
+        **kwargs
+    ) -> Tuple[List['BaseResource'], Dict]:
+        """
+        Extracts the resource from the kwargs and returns the resource 
+        and the remaining kwargs
+        """
+        resource_fields = get_pyd_field_names(resource)
+        resource_kwargs = {k: v for k, v in kwargs.items() if k in resource_fields}
+        return_kwargs = {k: v for k, v in kwargs.items() if k not in resource_fields}
+        resource_objs = []
+        for item in batch:
+            if isinstance(item, dict):
+                item.update(resource_kwargs)
+                resource_objs.append(resource.parse_obj(item))
+            else:
+                resource_objs.append(item)
+        return resource_objs, return_kwargs
+
     @classmethod
     def create_many(cls, data: List[Dict]) -> List['BaseResource']:
         """
@@ -267,7 +292,7 @@ class FileResource(BaseResource):
     file_id: Optional[str]
     filename: Optional[str] = None
     purpose: FilePurpose = FilePurpose.fine_tune
-    model: Optional[str]
+    model: Optional[str] = None
 
     @validator("purpose")
     def validate_purpose(cls, value):
@@ -299,4 +324,44 @@ async def async_get_params(self, **kwargs) -> List:
             files.append(
                 ("file", (self.filename or file.name, (await file.async_read_bytes() if _has_fileio else file.read_bytes()), "application/octet-stream"))
             )
-        return files
+        return files
+
+    @classmethod
+    def create_from_batch(
+        cls,
+        batch: List[Union[Dict[str, Any], str]],
+        output_path: Optional[str] = None,
+        file_id: Optional[str] = None,
+        filename: Optional[str] = None,
+        purpose: Optional[FilePurpose] = None,
+        **kwargs,
+    ) -> Tuple['FileObject', Dict[str, Any]]:
+        """
+        Creates a file object from a batch in jsonl format
+        """
+        for n, b in enumerate(batch):
+            if isinstance(b, dict):
+                batch[n] = json.dumps(b, cls = ObjectEncoder)
+        if output_path:
+            output = pathlib.Path(output_path)
+        else:
+            tmp = tempfile.NamedTemporaryFile(delete = False)
+            tmp.close()
+            output = pathlib.Path(tmp.name)
+
+        with output.open('w') as f:
+            for b in batch:
+                f.write(f'{b}\n')
+        resource_fields = get_pyd_field_names(cls)
+        resource_kwargs = {k: v for k, v in kwargs.items() if k in resource_fields}
+        return_kwargs = {k: v for k, v in kwargs.items() if k not in resource_fields}
+        return cls(
+            file = output,
+            purpose = purpose,
+            filename = filename,
+            file_id = file_id,
+            **resource_kwargs
+        ), return_kwargs
+        
+
+