Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,11 +1,22 @@
|
|
1 |
import gradio as gr
|
2 |
from huggingface_hub import InferenceClient
|
|
|
3 |
|
4 |
-
"""
|
5 |
-
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
|
6 |
-
"""
|
7 |
client = InferenceClient("Futuresony/future_ai_12_10_2024.gguf")
|
8 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
def respond(
|
11 |
message,
|
@@ -26,23 +37,30 @@ def respond(
|
|
26 |
messages.append({"role": "user", "content": message})
|
27 |
|
28 |
response = ""
|
29 |
-
|
30 |
-
|
31 |
messages,
|
32 |
max_tokens=max_tokens,
|
33 |
-
stream=True,
|
34 |
temperature=temperature,
|
35 |
top_p=top_p,
|
36 |
-
|
37 |
-
|
38 |
|
|
|
|
|
|
|
39 |
response += token
|
40 |
-
yield response
|
|
|
|
|
|
|
|
|
41 |
|
|
|
|
|
|
|
|
|
42 |
|
43 |
-
"""
|
44 |
-
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
|
45 |
-
"""
|
46 |
demo = gr.ChatInterface(
|
47 |
respond,
|
48 |
additional_inputs=[
|
@@ -59,6 +77,6 @@ demo = gr.ChatInterface(
|
|
59 |
],
|
60 |
)
|
61 |
|
62 |
-
|
63 |
if __name__ == "__main__":
|
64 |
demo.launch()
|
|
|
|
1 |
import gradio as gr
|
2 |
from huggingface_hub import InferenceClient
|
3 |
+
from web import search # Web search for fetching real-time answers
|
4 |
|
|
|
|
|
|
|
5 |
client = InferenceClient("Futuresony/future_ai_12_10_2024.gguf")
|
6 |
|
7 |
+
# Define topics that your model was NOT fine-tuned on
|
8 |
+
OUT_OF_SCOPE_TOPICS = ["current events", "latest news", "sports scores", "politics", "celebrity gossip"]
|
9 |
+
|
10 |
+
def is_out_of_scope(question):
|
11 |
+
"""Check if the question relates to topics the model wasn't trained on."""
|
12 |
+
return any(topic in question.lower() for topic in OUT_OF_SCOPE_TOPICS)
|
13 |
+
|
14 |
+
def google_search(query):
|
15 |
+
"""Fetch search results from Google."""
|
16 |
+
results = search(query)
|
17 |
+
if results:
|
18 |
+
return results[0] # Return the first result
|
19 |
+
return "Sorry, I couldn't find an answer on Google."
|
20 |
|
21 |
def respond(
|
22 |
message,
|
|
|
37 |
messages.append({"role": "user", "content": message})
|
38 |
|
39 |
response = ""
|
40 |
+
|
41 |
+
response_obj = client.chat_completion(
|
42 |
messages,
|
43 |
max_tokens=max_tokens,
|
|
|
44 |
temperature=temperature,
|
45 |
top_p=top_p,
|
46 |
+
logprobs=True, # Get log probabilities to measure uncertainty
|
47 |
+
)
|
48 |
|
49 |
+
# Extract response text
|
50 |
+
for message in response_obj:
|
51 |
+
token = message.choices[0].delta.content
|
52 |
response += token
|
53 |
+
yield response # Stream response
|
54 |
+
|
55 |
+
# Analyze uncertainty (using log probabilities)
|
56 |
+
logprobs = response_obj.choices[0].logprobs.token_logprobs
|
57 |
+
avg_confidence = sum(logprobs) / len(logprobs) if logprobs else 0
|
58 |
|
59 |
+
# If confidence is low OR the question is about out-of-scope topics, use Google
|
60 |
+
if avg_confidence < -5 or is_out_of_scope(message):
|
61 |
+
google_response = google_search(message)
|
62 |
+
yield f"🤖 AI (Low confidence): {response}\n\n🌍 Google: {google_response}"
|
63 |
|
|
|
|
|
|
|
64 |
demo = gr.ChatInterface(
|
65 |
respond,
|
66 |
additional_inputs=[
|
|
|
77 |
],
|
78 |
)
|
79 |
|
|
|
80 |
if __name__ == "__main__":
|
81 |
demo.launch()
|
82 |
+
|