add huggingface inference providers

app_huggingface.py

@@ -1,67 +1,20 @@
-import gradio as gr
-from gradio_client import Client, handle_file
+import ai_gradio
 
-MODELS = {"SmolVLM-Instruct": "akhaliq/SmolVLM-Instruct"}
+from utils_ai_gradio import get_app
 
+# Get the hyperbolic models but keep their full names for loading
+HUGGINGFACE_MODELS_FULL = [k for k in ai_gradio.registry.keys() if k.startswith("huggingface:")]
 
-def create_chat_fn(client):
-    def chat(message, history):
-        # Extract text and files from the message
-        text = message.get("text", "")
-        files = message.get("files", [])
+# Create display names without the prefix
+HUGGINGFACE_MODELS_DISPLAY = [k.replace("huggingface:", "") for k in HUGGINGFACE_MODELS_FULL]
 
-        # Handle file uploads if present
-        processed_files = [handle_file(f) for f in files]
 
-        response = client.predict(
-            message={"text": text, "files": processed_files},
-            system_prompt="You are a helpful AI assistant.",
-            temperature=0.7,
-            max_new_tokens=1024,
-            top_k=40,
-            repetition_penalty=1.1,
-            top_p=0.95,
-            api_name="/chat",
-        )
-        return response
-
-    return chat
-
-
-def set_client_for_session(model_name, request: gr.Request):
-    headers = {}
-    if request and hasattr(request, "headers"):
-        x_ip_token = request.headers.get("x-ip-token")
-        if x_ip_token:
-            headers["X-IP-Token"] = x_ip_token
-
-    return Client(MODELS[model_name], headers=headers)
-
-
-def safe_chat_fn(message, history, client):
-    if client is None:
-        return "Error: Client not initialized. Please refresh the page."
-    try:
-        return create_chat_fn(client)(message, history)
-    except Exception as e:
-        print(f"Error during chat: {e!s}")
-        return f"Error during chat: {e!s}"
-
-
-with gr.Blocks() as demo:
-    client = gr.State()
-
-    model_dropdown = gr.Dropdown(
-        choices=list(MODELS.keys()), value="SmolVLM-Instruct", label="Select Model", interactive=True
-    )
-
-    chat_interface = gr.ChatInterface(fn=safe_chat_fn, additional_inputs=[client], multimodal=True)
-
-    # Update client when model changes
-    model_dropdown.change(fn=set_client_for_session, inputs=[model_dropdown], outputs=[client])
-
-    # Initialize client on page load
-    demo.load(fn=set_client_for_session, inputs=[gr.State("SmolVLM-Instruct")], outputs=[client])
-
-if __name__ == "__main__":
-    demo.launch()
+# Create and launch the interface using get_app utility
+demo = get_app(
+    models=HUGGINGFACE_MODELS_FULL,  # Use the full names with prefix
+    default_model=HUGGINGFACE_MODELS_FULL[0],
+    dropdown_label="Select Huggingface Model",
+    choices=HUGGINGFACE_MODELS_DISPLAY,  # Display names without prefix
+    fill_height=True,
+    coder=True,
+)

pyproject.toml

@@ -38,7 +38,7 @@ dependencies = [
     "langchain>=0.3.14",
     "chromadb>=0.5.23",
     "openai>=1.55.0",
-    "ai-gradio[crewai,deepseek,gemini,groq,hyperbolic,openai,smolagents,transformers, langchain, mistral,minimax,nvidia, qwen, openrouter]>=0.2.46",
+    "ai-gradio[crewai,deepseek,gemini,groq,hyperbolic,openai,smolagents,transformers, langchain, mistral,minimax,nvidia, qwen, openrouter, huggingface]>=0.2.47",
 ]
 
 [tool.uv.sources]

requirements.txt

@@ -2,7 +2,7 @@
 #    uv pip compile pyproject.toml -o requirements.txt
 accelerate==1.2.1
     # via ai-gradio
-ai-gradio==0.2.46
+ai-gradio==0.2.47
     # via anychat (pyproject.toml)
 aiofiles==23.2.1
     # via gradio
@@ -428,9 +428,10 @@ httpx-sse==0.4.0
     #   langchain-community
 httpx-ws==0.7.1
     # via fireworks-ai
-huggingface-hub==0.27.1
+huggingface-hub==0.28.1
     # via
     #   accelerate
+    #   ai-gradio
     #   gradio
     #   gradio-client
     #   tokenizers