Using litellm instead of huggingfaceapi

app.py

@@ -2,7 +2,7 @@ import os
 import gradio as gr
 from llama_index.core.tools import FunctionTool
 from llama_index.core.agent.workflow import AgentWorkflow, ReActAgent
-from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
+from llama_index.llms.litellm import LiteLLM
 from code_agent import initialize_code_agent
 from scientific_paper_agent import load_scientific_paper_dataset, ScientificPaperRetriever
 from search_agent import init_search_tool
@@ -13,9 +13,9 @@ global currentMode
 
 hf_token = os.environ.get('HF_TOKEN')
 
-llm = HuggingFaceInferenceAPI(
-    model_name="Qwen/Qwen2.5-7B-Instruct",
-    token=hf_token
+llm = LiteLLM(
+    model_name="huggingface/Qwen/Qwen2.5-7B-Instruct",
+    api_key=hf_token
 )
 image_to_text_tool = FunctionTool.from_defaults(
     fn=init_image_to_text,
@@ -100,7 +100,7 @@ multi_agent_workflow = AgentWorkflow(
 
 
 
-def respond(
+async def respond(
     message,
     history: list[dict[str, str]],
     system_message,
@@ -110,7 +110,12 @@ def respond(
     mode
 ):
     global currentMode
-    print("Current Mode: " + mode)
+    if mode is not None:
+        mode = mode
+    else:
+        mode =  "Conversation Mode"
+
+    print(f"Current Mode: {mode}")
 
     if mode == "Math Mode":
         currentMode = "math"
@@ -121,7 +126,10 @@ def respond(
     else:
         currentMode = "conversation"
 
-    yield multi_agent_workflow.run(message, max_tokens=max_tokens, temperature=temperature, top_p=top_p)
+    result = await multi_agent_workflow.run(message, max_tokens=max_tokens, temperature=temperature, top_p=top_p)
+
+    return result
+
 
 
 with gr.Blocks() as demo:
@@ -131,11 +139,11 @@ with gr.Blocks() as demo:
         value="Conversation Mode",
         label="Interaction Mode"
     )
-
     # ChatInterface without additional_inputs
     chatbot = gr.ChatInterface(
         fn=respond,
-        type="messages"
+        type="messages",
+        additional_inputs=[mode_dropdown]
     )
 
     # Link dropdown value to respond function