add gpt-4o-2024-08-06 and json_schema for azure openAI service

langgenius · Aug 26, 2024 · dea6212 · dea6212
1 parent 7ae728a
commit dea6212
Show file tree

Hide file tree

Showing 2 changed files with 113 additions and 6 deletions.
diff --git a/api/core/model_runtime/model_providers/azure_openai/_constant.py b/api/core/model_runtime/model_providers/azure_openai/_constant.py
@@ -637,7 +637,19 @@ class AzureBaseModel(BaseModel):
                         en_US='specifying the format that the model must output'
                     ),
                     required=False,
-                    options=['text', 'json_object']
+                    options=['text', 'json_object', 'json_schema']
+                ),
+                ParameterRule(
+                    name='json_schema',
+                    label=I18nObject(
+                        en_US='JSON Schema'
+                    ),
+                    type='text',
+                    help=I18nObject(
+                        zh_Hans='设置返回的json schema，llm将按照它返回',
+                        en_US='Set a response json schema will ensure LLM to adhere it.'
+                    ),
+                    required=False
                 ),
             ],
             pricing=PriceConfig(
@@ -800,6 +812,94 @@ class AzureBaseModel(BaseModel):
             )
         )
     ),
+    AzureBaseModel(
+        base_model_name='gpt-4o-2024-08-06',
+        entity=AIModelEntity(
+            model='fake-deployment-name',
+            label=I18nObject(
+                en_US='fake-deployment-name-label',
+            ),
+            model_type=ModelType.LLM,
+            features=[
+                ModelFeature.AGENT_THOUGHT,
+                ModelFeature.VISION,
+                ModelFeature.MULTI_TOOL_CALL,
+                ModelFeature.STREAM_TOOL_CALL,
+            ],
+            fetch_from=FetchFrom.CUSTOMIZABLE_MODEL,
+            model_properties={
+                ModelPropertyKey.MODE: LLMMode.CHAT.value,
+                ModelPropertyKey.CONTEXT_SIZE: 128000,
+            },
+            parameter_rules=[
+                ParameterRule(
+                    name='temperature',
+                    **PARAMETER_RULE_TEMPLATE[DefaultParameterName.TEMPERATURE],
+                ),
+                ParameterRule(
+                    name='top_p',
+                    **PARAMETER_RULE_TEMPLATE[DefaultParameterName.TOP_P],
+                ),
+                ParameterRule(
+                    name='presence_penalty',
+                    **PARAMETER_RULE_TEMPLATE[DefaultParameterName.PRESENCE_PENALTY],
+                ),
+                ParameterRule(
+                    name='frequency_penalty',
+                    **PARAMETER_RULE_TEMPLATE[DefaultParameterName.FREQUENCY_PENALTY],
+                ),
+                _get_max_tokens(default=512, min_val=1, max_val=4096),
+                ParameterRule(
+                    name='seed',
+                    label=I18nObject(
+                        zh_Hans='种子',
+                        en_US='Seed'
+                    ),
+                    type='int',
+                    help=I18nObject(
+                        zh_Hans='如果指定，模型将尽最大努力进行确定性采样，使得重复的具有相同种子和参数的请求应该返回相同的结果。不能保证确定性，您应该参考 system_fingerprint 响应参数来监视变化。',
+                        en_US='If specified, model will make a best effort to sample deterministically, such that repeated requests with the same seed and parameters should return the same result. Determinism is not guaranteed, and you should refer to the system_fingerprint response parameter to monitor changes in the backend.'
+                    ),
+                    required=False,
+                    precision=2,
+                    min=0,
+                    max=1,
+                ),
+                ParameterRule(
+                    name='response_format',
+                    label=I18nObject(
+                        zh_Hans='回复格式',
+                        en_US='response_format'
+                    ),
+                    type='string',
+                    help=I18nObject(
+                        zh_Hans='指定模型必须输出的格式',
+                        en_US='specifying the format that the model must output'
+                    ),
+                    required=False,
+                    options=['text', 'json_object', 'json_schema']
+                ),
+                ParameterRule(
+                    name='json_schema',
+                    label=I18nObject(
+                        en_US='JSON Schema'
+                    ),
+                    type='text',
+                    help=I18nObject(
+                        zh_Hans='设置返回的json schema，llm将按照它返回',
+                        en_US='Set a response json schema will ensure LLM to adhere it.'
+                    ),
+                    required=False
+                ),
+            ],
+            pricing=PriceConfig(
+                input=5.00,
+                output=15.00,
+                unit=0.000001,
+                currency='USD',
+            )
+        )
+    ),
     AzureBaseModel(
         base_model_name='gpt-4-turbo',
         entity=AIModelEntity(

diff --git a/api/core/model_runtime/model_providers/azure_openai/llm/llm.py b/api/core/model_runtime/model_providers/azure_openai/llm/llm.py
@@ -1,4 +1,5 @@
 import copy
+import json
 import logging
 from collections.abc import Generator, Sequence
 from typing import Optional, Union, cast
@@ -276,12 +277,18 @@ def _chat_generate(self, model: str, credentials: dict,
 
         response_format = model_parameters.get("response_format")
         if response_format:
-            if response_format == "json_object":
-                response_format = {"type": "json_object"}
+            if response_format == "json_schema":
+                json_schema = model_parameters.get("json_schema")
+                if not json_schema:
+                    raise ValueError("Must define JSON Schema when the response format is json_schema")
+                try:
+                    schema = json.loads(json_schema)
+                except:
+                    raise ValueError(f"not currect json_schema format: {json_schema}")
+                model_parameters.pop("json_schema")
+                model_parameters["response_format"] = {"type": "json_schema", "json_schema": schema}
             else:
-                response_format = {"type": "text"}
-
-            model_parameters["response_format"] = response_format
+                model_parameters["response_format"] = {"type": response_format}
 
         extra_model_kwargs = {}