Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
from huggingface_hub import InferenceClient
|
|
|
4 |
|
5 |
class XylariaChat:
|
6 |
def __init__(self):
|
@@ -8,20 +9,22 @@ class XylariaChat:
|
|
8 |
self.hf_token = os.getenv("HF_TOKEN")
|
9 |
if not self.hf_token:
|
10 |
raise ValueError("HuggingFace token not found in environment variables")
|
11 |
-
|
12 |
# Initialize the inference client
|
13 |
self.client = InferenceClient(
|
14 |
-
model="Qwen/
|
15 |
api_key=self.hf_token
|
16 |
)
|
17 |
-
|
18 |
# Initialize conversation history and persistent memory
|
19 |
self.conversation_history = []
|
20 |
self.persistent_memory = {}
|
21 |
-
|
|
|
22 |
# System prompt with more detailed instructions
|
23 |
-
self.system_prompt = """You are a helpful and harmless AI assistant you are Xylaria 1.4 Senoa, Made by Sk Md Saad Amin you think step by step
|
24 |
"""
|
|
|
25 |
def store_information(self, key, value):
|
26 |
"""Store important information in persistent memory"""
|
27 |
self.persistent_memory[key] = value
|
@@ -30,25 +33,53 @@ class XylariaChat:
|
|
30 |
"""Retrieve information from persistent memory"""
|
31 |
return self.persistent_memory.get(key)
|
32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
def reset_conversation(self):
|
34 |
"""
|
35 |
-
Completely reset the conversation history, persistent memory,
|
36 |
and clear API-side memory
|
37 |
"""
|
38 |
# Clear local memory
|
39 |
self.conversation_history = []
|
40 |
self.persistent_memory.clear()
|
41 |
-
|
42 |
# Clear API-side memory by resetting the conversation
|
43 |
try:
|
44 |
# Attempt to clear any API-side session or context
|
45 |
self.client = InferenceClient(
|
46 |
-
model="Qwen/
|
47 |
api_key=self.hf_token
|
48 |
)
|
49 |
except Exception as e:
|
50 |
print(f"Error resetting API client: {e}")
|
51 |
-
|
|
|
52 |
return None # To clear the chatbot interface
|
53 |
|
54 |
def get_response(self, user_input):
|
@@ -58,14 +89,14 @@ class XylariaChat:
|
|
58 |
*self.conversation_history,
|
59 |
{"role": "user", "content": user_input}
|
60 |
]
|
61 |
-
|
62 |
# Add persistent memory context if available
|
63 |
if self.persistent_memory:
|
64 |
memory_context = "Remembered Information:\n" + "\n".join(
|
65 |
[f"{k}: {v}" for k, v in self.persistent_memory.items()]
|
66 |
)
|
67 |
messages.insert(1, {"role": "system", "content": memory_context})
|
68 |
-
|
69 |
# Generate response with streaming
|
70 |
try:
|
71 |
stream = self.client.chat.completions.create(
|
@@ -75,9 +106,9 @@ class XylariaChat:
|
|
75 |
top_p=0.7,
|
76 |
stream=True
|
77 |
)
|
78 |
-
|
79 |
return stream
|
80 |
-
|
81 |
except Exception as e:
|
82 |
return f"Error generating response: {str(e)}"
|
83 |
|
@@ -85,25 +116,25 @@ class XylariaChat:
|
|
85 |
def streaming_response(message, chat_history):
|
86 |
# Clear input textbox
|
87 |
response_stream = self.get_response(message)
|
88 |
-
|
89 |
# If it's an error, return immediately
|
90 |
if isinstance(response_stream, str):
|
91 |
return "", chat_history + [[message, response_stream]]
|
92 |
-
|
93 |
# Prepare for streaming response
|
94 |
full_response = ""
|
95 |
updated_history = chat_history + [[message, ""]]
|
96 |
-
|
97 |
# Streaming output
|
98 |
for chunk in response_stream:
|
99 |
if chunk.choices[0].delta.content:
|
100 |
chunk_content = chunk.choices[0].delta.content
|
101 |
full_response += chunk_content
|
102 |
-
|
103 |
# Update the last message in chat history with partial response
|
104 |
updated_history[-1][1] = full_response
|
105 |
yield "", updated_history
|
106 |
-
|
107 |
# Update conversation history
|
108 |
self.conversation_history.append(
|
109 |
{"role": "user", "content": message}
|
@@ -111,15 +142,22 @@ class XylariaChat:
|
|
111 |
self.conversation_history.append(
|
112 |
{"role": "assistant", "content": full_response}
|
113 |
)
|
114 |
-
|
115 |
# Limit conversation history to prevent token overflow
|
116 |
if len(self.conversation_history) > 10:
|
117 |
self.conversation_history = self.conversation_history[-10:]
|
118 |
|
119 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
120 |
custom_css = """
|
121 |
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&display=swap');
|
122 |
-
|
123 |
body, .gradio-container {
|
124 |
font-family: 'Inter', sans-serif !important;
|
125 |
}
|
@@ -133,64 +171,160 @@ class XylariaChat:
|
|
133 |
.gradio-container button {
|
134 |
font-family: 'Inter', sans-serif !important;
|
135 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
136 |
"""
|
137 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
138 |
with gr.Blocks(theme='soft', css=custom_css) as demo:
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
show_label=False,
|
151 |
-
placeholder="Type your message...",
|
152 |
-
container=False,
|
153 |
-
scale=4
|
154 |
)
|
155 |
-
btn = gr.Button("Send", scale=1)
|
156 |
-
|
157 |
-
# Clear history and memory buttons
|
158 |
-
clear = gr.Button("Clear Conversation")
|
159 |
-
clear_memory = gr.Button("Clear Memory")
|
160 |
-
|
161 |
-
# Submit functionality with streaming
|
162 |
-
btn.click(
|
163 |
-
fn=streaming_response,
|
164 |
-
inputs=[txt, chatbot],
|
165 |
-
outputs=[txt, chatbot]
|
166 |
-
)
|
167 |
-
txt.submit(
|
168 |
-
fn=streaming_response,
|
169 |
-
inputs=[txt, chatbot],
|
170 |
-
outputs=[txt, chatbot]
|
171 |
-
)
|
172 |
-
|
173 |
-
# Clear conversation history
|
174 |
-
clear.click(
|
175 |
-
fn=lambda: None,
|
176 |
-
inputs=None,
|
177 |
-
outputs=[chatbot],
|
178 |
-
queue=False
|
179 |
-
)
|
180 |
-
|
181 |
-
# Clear persistent memory and reset conversation
|
182 |
-
clear_memory.click(
|
183 |
-
fn=self.reset_conversation,
|
184 |
-
inputs=None,
|
185 |
-
outputs=[chatbot],
|
186 |
-
queue=False
|
187 |
-
)
|
188 |
|
189 |
-
|
190 |
-
|
191 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
192 |
return demo
|
193 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
194 |
# Launch the interface
|
195 |
def main():
|
196 |
chat = XylariaChat()
|
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
from huggingface_hub import InferenceClient
|
4 |
+
import json
|
5 |
|
6 |
class XylariaChat:
|
7 |
def __init__(self):
|
|
|
9 |
self.hf_token = os.getenv("HF_TOKEN")
|
10 |
if not self.hf_token:
|
11 |
raise ValueError("HuggingFace token not found in environment variables")
|
12 |
+
|
13 |
# Initialize the inference client
|
14 |
self.client = InferenceClient(
|
15 |
+
model="Qwen/Qwen-32B-Preview",
|
16 |
api_key=self.hf_token
|
17 |
)
|
18 |
+
|
19 |
# Initialize conversation history and persistent memory
|
20 |
self.conversation_history = []
|
21 |
self.persistent_memory = {}
|
22 |
+
self.chat_file_path = "chat_history.txt" # File to save chats
|
23 |
+
|
24 |
# System prompt with more detailed instructions
|
25 |
+
self.system_prompt = """You are a helpful and harmless AI assistant you are Xylaria 1.4 Senoa, Made by Sk Md Saad Amin you think step by step
|
26 |
"""
|
27 |
+
|
28 |
def store_information(self, key, value):
|
29 |
"""Store important information in persistent memory"""
|
30 |
self.persistent_memory[key] = value
|
|
|
33 |
"""Retrieve information from persistent memory"""
|
34 |
return self.persistent_memory.get(key)
|
35 |
|
36 |
+
def save_chat(self):
|
37 |
+
"""Saves the current chat history to a text file."""
|
38 |
+
try:
|
39 |
+
with open(self.chat_file_path, "w") as f:
|
40 |
+
chat_data = {
|
41 |
+
"conversation_history": self.conversation_history,
|
42 |
+
"persistent_memory": self.persistent_memory
|
43 |
+
}
|
44 |
+
json.dump(chat_data, f)
|
45 |
+
except Exception as e:
|
46 |
+
print(f"Error saving chat history: {e}")
|
47 |
+
|
48 |
+
def load_chat(self):
|
49 |
+
"""Loads chat history from a text file."""
|
50 |
+
try:
|
51 |
+
with open(self.chat_file_path, "r") as f:
|
52 |
+
chat_data = json.load(f)
|
53 |
+
self.conversation_history = chat_data.get("conversation_history", [])
|
54 |
+
self.persistent_memory = chat_data.get("persistent_memory", {})
|
55 |
+
return self.conversation_history, self.persistent_memory
|
56 |
+
except FileNotFoundError:
|
57 |
+
print("Chat history file not found.")
|
58 |
+
return [], {}
|
59 |
+
except Exception as e:
|
60 |
+
print(f"Error loading chat history: {e}")
|
61 |
+
return [], {}
|
62 |
+
|
63 |
def reset_conversation(self):
|
64 |
"""
|
65 |
+
Completely reset the conversation history, persistent memory,
|
66 |
and clear API-side memory
|
67 |
"""
|
68 |
# Clear local memory
|
69 |
self.conversation_history = []
|
70 |
self.persistent_memory.clear()
|
71 |
+
|
72 |
# Clear API-side memory by resetting the conversation
|
73 |
try:
|
74 |
# Attempt to clear any API-side session or context
|
75 |
self.client = InferenceClient(
|
76 |
+
model="Qwen/Qwen-32B-Preview",
|
77 |
api_key=self.hf_token
|
78 |
)
|
79 |
except Exception as e:
|
80 |
print(f"Error resetting API client: {e}")
|
81 |
+
|
82 |
+
self.save_chat() # Save the empty chat history
|
83 |
return None # To clear the chatbot interface
|
84 |
|
85 |
def get_response(self, user_input):
|
|
|
89 |
*self.conversation_history,
|
90 |
{"role": "user", "content": user_input}
|
91 |
]
|
92 |
+
|
93 |
# Add persistent memory context if available
|
94 |
if self.persistent_memory:
|
95 |
memory_context = "Remembered Information:\n" + "\n".join(
|
96 |
[f"{k}: {v}" for k, v in self.persistent_memory.items()]
|
97 |
)
|
98 |
messages.insert(1, {"role": "system", "content": memory_context})
|
99 |
+
|
100 |
# Generate response with streaming
|
101 |
try:
|
102 |
stream = self.client.chat.completions.create(
|
|
|
106 |
top_p=0.7,
|
107 |
stream=True
|
108 |
)
|
109 |
+
|
110 |
return stream
|
111 |
+
|
112 |
except Exception as e:
|
113 |
return f"Error generating response: {str(e)}"
|
114 |
|
|
|
116 |
def streaming_response(message, chat_history):
|
117 |
# Clear input textbox
|
118 |
response_stream = self.get_response(message)
|
119 |
+
|
120 |
# If it's an error, return immediately
|
121 |
if isinstance(response_stream, str):
|
122 |
return "", chat_history + [[message, response_stream]]
|
123 |
+
|
124 |
# Prepare for streaming response
|
125 |
full_response = ""
|
126 |
updated_history = chat_history + [[message, ""]]
|
127 |
+
|
128 |
# Streaming output
|
129 |
for chunk in response_stream:
|
130 |
if chunk.choices[0].delta.content:
|
131 |
chunk_content = chunk.choices[0].delta.content
|
132 |
full_response += chunk_content
|
133 |
+
|
134 |
# Update the last message in chat history with partial response
|
135 |
updated_history[-1][1] = full_response
|
136 |
yield "", updated_history
|
137 |
+
|
138 |
# Update conversation history
|
139 |
self.conversation_history.append(
|
140 |
{"role": "user", "content": message}
|
|
|
142 |
self.conversation_history.append(
|
143 |
{"role": "assistant", "content": full_response}
|
144 |
)
|
145 |
+
|
146 |
# Limit conversation history to prevent token overflow
|
147 |
if len(self.conversation_history) > 10:
|
148 |
self.conversation_history = self.conversation_history[-10:]
|
149 |
|
150 |
+
self.save_chat()
|
151 |
+
|
152 |
+
def load_chat_interface():
|
153 |
+
"""Loads the chat history into the chatbot interface."""
|
154 |
+
self.load_chat()
|
155 |
+
return self.conversation_history
|
156 |
+
|
157 |
+
# Custom CSS for Inter font and sidebar
|
158 |
custom_css = """
|
159 |
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&display=swap');
|
160 |
+
|
161 |
body, .gradio-container {
|
162 |
font-family: 'Inter', sans-serif !important;
|
163 |
}
|
|
|
171 |
.gradio-container button {
|
172 |
font-family: 'Inter', sans-serif !important;
|
173 |
}
|
174 |
+
|
175 |
+
/* Sidebar styling */
|
176 |
+
#sidebar {
|
177 |
+
background-color: #f2f2f2;
|
178 |
+
border-right: 1px solid #ccc;
|
179 |
+
padding: 10px;
|
180 |
+
height: 100vh;
|
181 |
+
overflow-y: auto;
|
182 |
+
}
|
183 |
+
|
184 |
+
#sidebar ul {
|
185 |
+
list-style-type: none;
|
186 |
+
padding: 0;
|
187 |
+
}
|
188 |
+
|
189 |
+
#sidebar li {
|
190 |
+
margin-bottom: 5px;
|
191 |
+
}
|
192 |
+
/* Main chat area */
|
193 |
+
#main-chat {
|
194 |
+
padding: 20px;
|
195 |
+
}
|
196 |
"""
|
197 |
|
198 |
+
# Example prompts
|
199 |
+
example_prompts = [
|
200 |
+
"How do I get started with coding?",
|
201 |
+
"Tell me a fun fact about science.",
|
202 |
+
"What are some good books to read?"
|
203 |
+
]
|
204 |
+
|
205 |
+
# Function to forward prompt to the textbox
|
206 |
+
def forward_prompt(prompt):
|
207 |
+
return prompt
|
208 |
+
|
209 |
with gr.Blocks(theme='soft', css=custom_css) as demo:
|
210 |
+
with gr.Row():
|
211 |
+
# Sidebar for displaying chat history
|
212 |
+
with gr.Column(elem_id="sidebar", scale=1): # Added elem_id for CSS
|
213 |
+
gr.Markdown("### Chat History")
|
214 |
+
load_button = gr.Button("Load Chat History")
|
215 |
+
chat_list = gr.Markdown("No chat history found.")
|
216 |
+
|
217 |
+
load_button.click(
|
218 |
+
fn=lambda: gr.Markdown.update(value=self.format_chat_history()),
|
219 |
+
inputs=None,
|
220 |
+
outputs=[chat_list]
|
|
|
|
|
|
|
|
|
221 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
222 |
|
223 |
+
# Main chat interface
|
224 |
+
with gr.Column(elem_id="main-chat", scale=3):
|
225 |
+
# Show Xylaria and example prompts only on the first page/new chat
|
226 |
+
with gr.Column(visible=True) as start_page:
|
227 |
+
gr.Markdown("# Xylaria")
|
228 |
+
with gr.Row():
|
229 |
+
for prompt in example_prompts:
|
230 |
+
gr.Button(prompt).click(
|
231 |
+
fn=forward_prompt,
|
232 |
+
inputs=gr.State(prompt),
|
233 |
+
outputs=txt
|
234 |
+
)
|
235 |
+
|
236 |
+
with gr.Column(visible=False) as chat_page:
|
237 |
+
chatbot = gr.Chatbot(
|
238 |
+
label="Xylaria 1.4 Senoa",
|
239 |
+
height=500,
|
240 |
+
show_copy_button=True
|
241 |
+
)
|
242 |
+
|
243 |
+
# Input row with improved layout
|
244 |
+
with gr.Row():
|
245 |
+
txt = gr.Textbox(
|
246 |
+
show_label=False,
|
247 |
+
placeholder="Type your message...",
|
248 |
+
container=False,
|
249 |
+
scale=4
|
250 |
+
)
|
251 |
+
btn = gr.Button("Send", scale=1)
|
252 |
+
|
253 |
+
# Clear history and memory buttons
|
254 |
+
clear = gr.Button("Clear Conversation")
|
255 |
+
clear_memory = gr.Button("Clear Memory")
|
256 |
+
|
257 |
+
# Toggle between start page and chat page
|
258 |
+
def toggle_page(choice):
|
259 |
+
return gr.Column.update(visible=choice == "chat"), gr.Column.update(visible=choice == "start")
|
260 |
+
|
261 |
+
start_page.visible = True
|
262 |
+
chat_page.visible = False
|
263 |
+
|
264 |
+
# Submit functionality with streaming
|
265 |
+
btn.click(
|
266 |
+
fn=streaming_response,
|
267 |
+
inputs=[txt, chatbot],
|
268 |
+
outputs=[txt, chatbot]
|
269 |
+
).then(
|
270 |
+
fn=lambda: toggle_page("chat"),
|
271 |
+
inputs=gr.State("chat"),
|
272 |
+
outputs=[chat_page, start_page]
|
273 |
+
)
|
274 |
+
txt.submit(
|
275 |
+
fn=streaming_response,
|
276 |
+
inputs=[txt, chatbot],
|
277 |
+
outputs=[txt, chatbot]
|
278 |
+
).then(
|
279 |
+
fn=lambda: toggle_page("chat"),
|
280 |
+
inputs=gr.State("chat"),
|
281 |
+
outputs=[chat_page, start_page]
|
282 |
+
)
|
283 |
+
|
284 |
+
# Clear conversation history
|
285 |
+
clear.click(
|
286 |
+
fn=lambda: None,
|
287 |
+
inputs=None,
|
288 |
+
outputs=[chatbot],
|
289 |
+
queue=False
|
290 |
+
).then(
|
291 |
+
fn=lambda: toggle_page("start"),
|
292 |
+
inputs=gr.State("start"),
|
293 |
+
outputs=[chat_page, start_page]
|
294 |
+
)
|
295 |
+
|
296 |
+
# Clear persistent memory and reset conversation
|
297 |
+
clear_memory.click(
|
298 |
+
fn=self.reset_conversation,
|
299 |
+
inputs=None,
|
300 |
+
outputs=[chatbot],
|
301 |
+
queue=False
|
302 |
+
).then(
|
303 |
+
fn=lambda: toggle_page("start"),
|
304 |
+
inputs=gr.State("start"),
|
305 |
+
outputs=[chat_page, start_page]
|
306 |
+
)
|
307 |
+
|
308 |
+
# Ensure memory is cleared when the interface is closed
|
309 |
+
demo.load(self.reset_conversation, None, None)
|
310 |
+
|
311 |
return demo
|
312 |
|
313 |
+
def format_chat_history(self):
|
314 |
+
"""Formats the chat history for display in the sidebar."""
|
315 |
+
self.load_chat() # Load the chat history first
|
316 |
+
if not self.conversation_history:
|
317 |
+
return "No chat history found."
|
318 |
+
|
319 |
+
formatted_history = ""
|
320 |
+
for chat in self.conversation_history:
|
321 |
+
if chat["role"] == "user":
|
322 |
+
formatted_history += f"**You:** {chat['content']}\n\n"
|
323 |
+
elif chat["role"] == "assistant":
|
324 |
+
formatted_history += f"**Xylaria:** {chat['content']}\n\n"
|
325 |
+
|
326 |
+
return formatted_history
|
327 |
+
|
328 |
# Launch the interface
|
329 |
def main():
|
330 |
chat = XylariaChat()
|