feat: add qianfan llm provider

Author: jimmyzhuu

.env-template

@@ -9,6 +9,7 @@ INTERNAL_KEY=<internal key for worker-to-backend authentication>
 # GOOGLE_API_KEY=<your-google-api-key>
 # GROQ_API_KEY=<your-groq-api-key>
 # NOVITA_API_KEY=<your-novita-api-key>
+# QIANFAN_API_KEY=<your-qianfan-api-key>
 # OPEN_ROUTER_API_KEY=<your-openrouter-api-key>
 
 # Remote Embeddings (Optional - for using a remote embeddings API instead of local SentenceTransformer)

application/core/model_configs.py

@@ -28,6 +28,7 @@
 OPENROUTER_ATTACHMENTS = IMAGE_ATTACHMENTS
 
 NOVITA_ATTACHMENTS = IMAGE_ATTACHMENTS
+QIANFAN_ATTACHMENTS = []
 
 
 OPENAI_MODELS = [
@@ -234,6 +235,21 @@
     ),
 ]
 
+QIANFAN_MODELS = [
+    AvailableModel(
+        id="ernie-5.0",
+        provider=ModelProvider.QIANFAN,
+        display_name="ERNIE 5.0",
+        description="Baidu Qianfan flagship text chat model",
+        capabilities=ModelCapabilities(
+            supports_tools=False,
+            supports_structured_output=False,
+            supported_attachment_types=QIANFAN_ATTACHMENTS,
+            context_window=128000,
+        ),
+    ),
+]
+
 
 AZURE_OPENAI_MODELS = [
     AvailableModel(

application/core/model_settings.py

@@ -19,6 +19,7 @@ class ModelProvider(str, Enum):
     PREMAI = "premai"
     SAGEMAKER = "sagemaker"
     NOVITA = "novita"
+    QIANFAN = "qianfan"
 
 
 @dataclass
@@ -118,6 +119,10 @@ def _load_models(self):
             settings.LLM_PROVIDER == "novita" and settings.API_KEY
         ):
             self._add_novita_models(settings)
+        if settings.QIANFAN_API_KEY or (
+            settings.LLM_PROVIDER == "qianfan" and settings.API_KEY
+        ):
+            self._add_qianfan_models(settings)
         if settings.HUGGINGFACE_API_KEY or (
             settings.LLM_PROVIDER == "huggingface" and settings.API_KEY
         ):
@@ -264,6 +269,21 @@ def _add_novita_models(self, settings):
         for model in NOVITA_MODELS:
             self.models[model.id] = model
 
+    def _add_qianfan_models(self, settings):
+        from application.core.model_configs import QIANFAN_MODELS
+
+        if settings.QIANFAN_API_KEY:
+            for model in QIANFAN_MODELS:
+                self.models[model.id] = model
+            return
+        if settings.LLM_PROVIDER == "qianfan" and settings.LLM_NAME:
+            for model in QIANFAN_MODELS:
+                if model.id == settings.LLM_NAME:
+                    self.models[model.id] = model
+                    return
+        for model in QIANFAN_MODELS:
+            self.models[model.id] = model
+
     def _add_docsgpt_models(self, settings):
         model_id = "docsgpt-local"
         model = AvailableModel(

application/core/model_utils.py

@@ -11,6 +11,7 @@ def get_api_key_for_provider(provider: str) -> Optional[str]:
         "openai": settings.OPENAI_API_KEY,
         "openrouter": settings.OPEN_ROUTER_API_KEY,
         "novita": settings.NOVITA_API_KEY,
+        "qianfan": settings.QIANFAN_API_KEY,
         "anthropic": settings.ANTHROPIC_API_KEY,
         "google": settings.GOOGLE_API_KEY,
         "groq": settings.GROQ_API_KEY,

application/core/settings.py

@@ -79,6 +79,7 @@ class Settings(BaseSettings):
     HUGGINGFACE_API_KEY: Optional[str] = None
     OPEN_ROUTER_API_KEY: Optional[str] = None
     NOVITA_API_KEY: Optional[str] = None
+    QIANFAN_API_KEY: Optional[str] = None
 
     OPENAI_API_BASE: Optional[str] = None  # azure openai api base url
     OPENAI_API_VERSION: Optional[str] = None  # azure openai api version
@@ -164,6 +165,7 @@ class Settings(BaseSettings):
         "GROQ_API_KEY",
         "HUGGINGFACE_API_KEY",
         "NOVITA_API_KEY",
+        "QIANFAN_API_KEY",
         "EMBEDDINGS_KEY",
         "FALLBACK_LLM_API_KEY",
         "QDRANT_API_KEY",

application/llm/handlers/handler_creator.py

@@ -8,6 +8,7 @@ class LLMHandlerCreator:
         "openai": OpenAILLMHandler,
         "google": GoogleLLMHandler,
         "novita": OpenAILLMHandler,  # Novita uses OpenAI-compatible API
+        "qianfan": OpenAILLMHandler,  # Qianfan uses an OpenAI-compatible chat API
         "default": OpenAILLMHandler,
     }
 

application/llm/llm_creator.py

@@ -7,6 +7,7 @@
 from application.llm.llama_cpp import LlamaCpp
 from application.llm.novita import NovitaLLM
 from application.llm.openai import AzureOpenAILLM, OpenAILLM
+from application.llm.qianfan import QianfanLLM
 from application.llm.premai import PremAILLM
 from application.llm.sagemaker import SagemakerAPILLM
 from application.llm.open_router import OpenRouterLLM
@@ -26,6 +27,7 @@ class LLMCreator:
         "groq": GroqLLM,
         "google": GoogleLLM,
         "novita": NovitaLLM,
+        "qianfan": QianfanLLM,
         "openrouter": OpenRouterLLM,
     }
 

application/llm/qianfan.py

@@ -0,0 +1,27 @@
+from application.core.settings import settings
+from application.llm.openai import OpenAILLM
+
+QIANFAN_BASE_URL = "https://qianfan.baidubce.com/v2"
+
+
+class QianfanLLM(OpenAILLM):
+    def __init__(self, api_key=None, user_api_key=None, base_url=None, *args, **kwargs):
+        super().__init__(
+            api_key=api_key or settings.QIANFAN_API_KEY or settings.API_KEY,
+            user_api_key=user_api_key,
+            base_url=base_url or QIANFAN_BASE_URL,
+            *args,
+            **kwargs,
+        )
+
+    def get_supported_attachment_types(self):
+        """Keep the first Qianfan integration text-only until attachment support is verified."""
+        return []
+
+    def _supports_tools(self):
+        """Disable tools until Qianfan tool-calling is verified end-to-end."""
+        return False
+
+    def _supports_structured_output(self):
+        """Disable structured output until JSON schema support is verified end-to-end."""
+        return False

application/utils.py

@@ -35,6 +35,7 @@ def get_gpt_model() -> str:
         "anthropic": "claude-2",
         "groq": "llama3-8b-8192",
         "novita": "deepseek/deepseek-r1",
+        "qianfan": "ernie-5.0",
     }
     return settings.LLM_NAME or model_map.get(settings.LLM_PROVIDER, "")
 

docs/content/Deploying/DocsGPT-Settings.mdx

@@ -45,6 +45,7 @@ Here are some of the most fundamental settings you'll likely want to configure:
     - `google`: Use Google's Vertex AI or Gemini models.
     - `anthropic`: Use Anthropic's Claude models.
     - `groq`: Use Groq's models.
+    - `qianfan`: Use Baidu Qianfan models.
     - `huggingface`: Use HuggingFace Inference API.
     - `azure_openai`: Use Azure OpenAI Service.
     - `openai` (when using local inference engines like Ollama, Llama.cpp, TGI, etc.): This signals DocsGPT to use an OpenAI-compatible API format, even if the actual LLM is running locally.
@@ -54,8 +55,26 @@ Here are some of the most fundamental settings you'll likely want to configure:
   - **Examples:**
     - For `LLM_PROVIDER=openai`: `gpt-4o`
     - For `LLM_PROVIDER=google`: `gemini-2.0-flash`
+    - For `LLM_PROVIDER=qianfan`: `ernie-5.0`
     - For local models (e.g., Ollama): `llama3.2:1b` (or any model name available in your setup).
 
+### Example for Cloud API Provider (Baidu Qianfan)
+
+To use Baidu Qianfan's `ernie-5.0` model, configure your `.env` file like this:
+
+```env
+LLM_PROVIDER=qianfan
+QIANFAN_API_KEY=YOUR_QIANFAN_API_KEY
+LLM_NAME=ernie-5.0
+```
+
+You can also reuse the generic `API_KEY` variable with `LLM_PROVIDER=qianfan`, but `QIANFAN_API_KEY` is clearer when you configure multiple providers in the same deployment.
+
+The first Qianfan integration is intentionally conservative:
+
+- Text chat is supported.
+- Attachments, tool calling, and structured output are left disabled until they are verified end-to-end against the Qianfan API.
+
 - **`EMBEDDINGS_NAME`**: This setting defines which embedding model DocsGPT will use to generate vector embeddings for your documents. Embeddings are numerical representations of text that allow DocsGPT to understand the semantic meaning of your documents for efficient search and retrieval.
 
   - **Default value:** `huggingface_sentence-transformers/all-mpnet-base-v2` (a good general-purpose embedding model).