feat(model): API support reasoning (#2409)

packages/dbgpt-client/src/dbgpt_client/_cli.py

@@ -245,7 +245,7 @@ def _run_flow_cmd_local(
                 if not out.success:
                     cl.error(out.text)
                 else:
-                    cl.print(out.text, end="")
+                    cl.print(out.gen_text_with_thinking(), end="")
         except Exception as e:
             cl.error(f"Failed to run flow: {e}", exit_code=1)
         finally:
@@ -436,7 +436,7 @@ def _run_flow_chat_local(
                 cl.error(f"Error: {out.text}")
                 raise Exception(out.text)
             else:
-                yield out.text
+                yield out.gen_text_with_thinking()
 
     async def _call(_call_body: Dict[str, Any]):
         nonlocal dag, dag_metadata
@@ -466,6 +466,9 @@ def _run_flow_chat_stream(
         async for out in client.chat_stream(**_call_body):
             if out.choices:
                 text = out.choices[0].delta.content
+                reasoning_content = out.choices[0].delta.reasoning_content
+                if reasoning_content:
+                    yield reasoning_content
                 if text:
                     yield text
 
@@ -482,6 +485,13 @@ def _run_flow_chat(
         res = await client.chat(**_call_body)
         if res.choices:
             text = res.choices[0].message.content
+            if res.choices[0].message.reasoning_content:
+                reasoning_content = res.choices[0].message.reasoning_content
+                # For each line, add '>' at the beginning
+                reasoning_content = "\n".join(
+                    [f"> {line}" for line in reasoning_content.split("\n")]
+                )
+                text = reasoning_content + "\n\n" + text
             return text
 
     loop.run_until_complete(_chat(_call, interactive, json_data))

packages/dbgpt-client/src/dbgpt_client/schema.py

@@ -7,32 +7,20 @@ from typing import Any, Dict, List, Optional, Union
 
 from fastapi import File, UploadFile
 
-from dbgpt._private.pydantic import BaseModel, ConfigDict, Field
+from dbgpt._private.pydantic import BaseModel, ConfigDict, Field, model_validator
+from dbgpt.core.awel import CommonLLMHttpRequestBody
+from dbgpt.core.schema.api import APIChatCompletionRequest
 from dbgpt_ext.rag.chunk_manager import ChunkParameters
 
 
-class ChatCompletionRequestBody(BaseModel):
+class ChatCompletionRequestBody(APIChatCompletionRequest):
     """ChatCompletion LLM http request body."""
 
-    model: str = Field(
-        ..., description="The model name", examples=["gpt-3.5-turbo", "proxyllm"]
-    )
-    messages: Union[str, List[str]] = Field(
-        ..., description="User input messages", examples=["Hello", "How are you?"]
-    )
-    stream: bool = Field(default=True, description="Whether return stream")
-
-    temperature: Optional[float] = Field(
-        default=None,
-        description="What sampling temperature to use, between 0 and 2. Higher values "
-        "like 0.8 will make the output more random, "
-        "while lower values like 0.2 will "
-        "make it more focused and deterministic.",
-    )
     max_new_tokens: Optional[int] = Field(
         default=None,
         description="The maximum number of tokens that can be generated in the chat "
         "completion.",
+        deprecated="'max_new_tokens' is deprecated. Use 'max_tokens' instead.",
     )
     conv_uid: Optional[str] = Field(
         default=None, description="The conversation id of the model inference"
@@ -65,6 +53,37 @@ class ChatCompletionRequestBody(BaseModel):
         default=True, description="response content whether to output vis label"
     )
 
+    @model_validator(mode="before")
+    @classmethod
+    def pre_fill(cls, values: Dict[str, Any]) -> Dict[str, Any]:
+        """Pre fill the messages."""
+        if not isinstance(values, dict):
+            return values
+        max_tokens = values.get("max_tokens")
+        max_new_tokens = values.get("max_new_tokens")
+        if max_tokens is None and max_new_tokens is not None:
+            values["max_tokens"] = max_new_tokens
+        return values
+
+    def to_common_llm_http_request_body(self) -> CommonLLMHttpRequestBody:
+        """Convert to CommonLLMHttpRequestBody."""
+        max_new_tokens = self.max_tokens
+        return CommonLLMHttpRequestBody(
+            model=self.model,
+            messages=self.single_prompt(),
+            stream=self.stream,
+            temperature=self.temperature,
+            max_new_tokens=max_new_tokens,
+            conv_uid=self.conv_uid,
+            span_id=self.span_id,
+            chat_mode=self.chat_mode,
+            chat_param=self.chat_param,
+            user_name=self.user_name,
+            sys_code=self.sys_code,
+            incremental=self.incremental,
+            enable_vis=self.enable_vis,
+        )
+
 
 class ChatMode(Enum):
     """Chat mode."""
@@ -74,6 +93,7 @@ class ChatMode(Enum):
     CHAT_AWEL_FLOW = "chat_flow"
     CHAT_KNOWLEDGE = "chat_knowledge"
     CHAT_DATA = "chat_data"
+    CHAT_DB_QA = "chat_with_db_qa"
 
 
 class AWELTeamModel(BaseModel):