Update app.py

app.py

@@ -20,6 +20,8 @@ from sympy.parsing.latex import parse_latex
 import base64
 from transformers import AutoModelForSequenceClassification, AutoTokenizer
 from transformers import AutoTokenizer, AutoModelForPreTraining
+from langchain_community.llms.manifest import ManifestWrapper
+from manifest import Manifest
 
 #client = OpenAI(
    # base_url=os.environ.get("SERVER_URL"),
@@ -534,36 +536,21 @@ def parse_data_chunk(data_chunk):
         data_chunk = data_chunk.text
     chunks = data_chunk.split("data:")
 
-def generate(message, temperature):
+def parse_data_chunk(data_chunk):
     """
-    Generates a chat completion response by streaming data from the client chat model.
-
-    This function streams the response from the client chat model and yields the content
-    of the response chunk by chunk. If an error occurs, it yields the error message.
-
+    Parse a given data chunk string into a list of individual data entries.
+    The function splits the input string by the delimiter "data:" and removes any
+    leading or trailing whitespace from each resulting chunk. Empty chunks are
+    filtered out from the final list.
     Parameters:
-    message (str): The input message to be sent to the chat model.
-    temperature (float): The sampling temperature to use. Higher values mean the model will take more risks.
-
-    Yields:
-    tuple: A tuple containing the content of the response and a boolean flag indicating if an error occurred.
-           If no error occurred, the boolean flag will be False and the content will be the response text.
-           If an error occurred, the boolean flag will be True and the content will be the error message.
+    data_chunk (str): The input string containing data chunks separated by "data:".
+    Returns:
+    list: A list of individual data entries with whitespace stripped.
     """
-    response = client.chat.completions.create(
-        model="H4/zephyr-7b-beta",
-        messages=message,
-        stream=True,
-        max_tokens=1024,
-        stop=["output\n"],
-        temperature=temperature,
-        #timeout=30,
-    )
-
-    #response = stream.text
-
-    # The reason why the library method is not used here is that if an error occurs,
-    #    the returned data will not be a stream, and using the official library will result in an error.
+    if isinstance(data_chunk, InferenceClient.ChatCompletionStreamOutput):  # Update this line if you're using a different client class
+        data_chunk = data_chunk.text
+    chunks = data_chunk.split("data:")
+
     for chunk in response:
         chunk = chunk.decode("utf-8") if isinstance(chunk, bytes) else chunk
         data_chunks = parse_data_chunk(chunk)