conntect to urzedasek endpoint

.gitignore

@@ -0,0 +1,2 @@
+.env
+.idea

app.py

@@ -1,21 +1,46 @@
+import time
+import os
 import gradio as gr
-from huggingface_hub import InferenceClient
+import requests as requests
+
+
+URZEDASEK_ENDPOINT = os.getenv("URZEDASEK_ENDPOINT", "http://0.0.0.0:8000/v1/chat/urzedasek")
+ANONYMOUS_TOKEN = os.getenv("ANONYMOUS_TOKEN")
+
+
+class UrzedasekInferenceClient():
+
+    def chat_completion(
+        self,
+        messages,
+    ):
+        auth_header = {
+            "Authorization": f"Token {ANONYMOUS_TOKEN}"
+            }
+        ret = requests.post(
+            URZEDASEK_ENDPOINT,
+            headers=auth_header,
+            json={
+                "messages": messages,
+            },
+        )
+        return ret.json()['result']['urzedas_answer_builder']['replies']
+
+
+client = UrzedasekInferenceClient()
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+
+def chunker(result):
+    for i in range(0, len(result), 40):
+        time.sleep(0.1)
+        yield result[i:i + 40]
 
 
 def respond(
     message,
     history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
 ):
-    messages = [{"role": "system", "content": system_message}]
+    messages = []
 
     for val in history:
         if val[0]:
@@ -27,17 +52,11 @@ def respond(
 
     response = ""
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+    for message in client.chat_completion(messages,):
+        for chunk in chunker(message):
+            response += chunk
+            yield response
 
-        response += token
-        yield response
 
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
@@ -45,19 +64,10 @@ For information on how to customize the ChatInterface, peruse the gradio docs: h
 demo = gr.ChatInterface(
     respond,
     additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+        # gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
     ],
 )
 
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(share=True)

requirements.txt

@@ -1 +1,2 @@
-huggingface_hub==0.22.2
+huggingface_hub==0.22.2
+requests==2.31.0