Spaces:
Running
on
Zero
Running
on
Zero
eaglesarezzo
commited on
Commit
โข
3f057a2
1
Parent(s):
e67fd61
Update app.py
Browse files
app.py
CHANGED
@@ -24,15 +24,6 @@ backend = Backend()
|
|
24 |
|
25 |
cv2.setNumThreads(1)
|
26 |
|
27 |
-
def load_topic_data(topic):
|
28 |
-
if topic:
|
29 |
-
query_engine = backend.create_index_for_query_engine(documents_paths[topic])
|
30 |
-
gr.Info(f"Data loaded for {topic} topic")
|
31 |
-
return query_engine
|
32 |
-
else:
|
33 |
-
gr.Warning("No topic selected. Please select a topic first.")
|
34 |
-
return None
|
35 |
-
|
36 |
@spaces.GPU(duration=20)
|
37 |
def respond(
|
38 |
message,
|
@@ -44,14 +35,21 @@ def respond(
|
|
44 |
top_p,
|
45 |
top_k,
|
46 |
repeat_penalty,
|
47 |
-
|
48 |
):
|
49 |
chat_template = MessagesFormatterType.GEMMA_2
|
50 |
|
51 |
-
|
52 |
-
|
53 |
|
54 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
|
56 |
# Load model only if it's not already loaded or if a new model is selected
|
57 |
if backend.llm is None or backend.llm_model != model:
|
@@ -96,21 +94,12 @@ def respond(
|
|
96 |
outputs = ""
|
97 |
for output in stream:
|
98 |
outputs += output
|
99 |
-
yield history + [[message, outputs]]
|
100 |
except Exception as e:
|
101 |
yield history + [[message, f"Error during response generation: {str(e)}"]]
|
102 |
|
103 |
def select_topic(topic):
|
104 |
-
|
105 |
-
return (
|
106 |
-
gr.update(interactive=True), # Enable the chat input
|
107 |
-
gr.update(interactive=True), # Enable the submit button
|
108 |
-
gr.update(visible=True), # Make the chatbot visible
|
109 |
-
gr.update(interactive=False), # Disable blockchain button
|
110 |
-
gr.update(interactive=False), # Disable metaverse button
|
111 |
-
gr.update(interactive=False), # Disable payment button
|
112 |
-
query_engine # Return the loaded query engine
|
113 |
-
)
|
114 |
|
115 |
with gr.Blocks(css="""
|
116 |
.gradio-container {
|
@@ -125,7 +114,7 @@ with gr.Blocks(css="""
|
|
125 |
metaverse_btn = gr.Button("๐ Metaverse", scale=1)
|
126 |
payment_btn = gr.Button("๐ณ Payment", scale=1)
|
127 |
|
128 |
-
|
129 |
|
130 |
chatbot = gr.Chatbot(
|
131 |
scale=1,
|
@@ -140,9 +129,8 @@ with gr.Blocks(css="""
|
|
140 |
show_label=False,
|
141 |
placeholder="Inserisci il tuo messaggio...",
|
142 |
container=False,
|
143 |
-
interactive=False
|
144 |
)
|
145 |
-
submit_btn = gr.Button("Invia", scale=1
|
146 |
|
147 |
with gr.Accordion("Advanced Options", open=False):
|
148 |
model = gr.Dropdown([
|
@@ -158,29 +146,26 @@ with gr.Blocks(css="""
|
|
158 |
value="gemma-2-2b-it-Q6_K_L.gguf",
|
159 |
label="Model"
|
160 |
)
|
161 |
-
system_message = gr.Textbox(value="""
|
162 |
-
Nel fornire la risposta cita il report da cui la hai ottenuta.
|
163 |
-
Utilizza la cronologia della chat o il contesto fornito per aiutare l'utente a ottenere una risposta accurata.
|
164 |
-
Non rispondere mai a domande che non sono pertinenti a questi argomenti.""", label="System message")
|
165 |
max_tokens = gr.Slider(minimum=1, maximum=4096, value=3048, step=1, label="Max tokens")
|
166 |
temperature = gr.Slider(minimum=0.1, maximum=4.0, value=1.2, step=0.1, label="Temperature")
|
167 |
top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p")
|
168 |
top_k = gr.Slider(minimum=0, maximum=100, value=30, step=1, label="Top-k")
|
169 |
repeat_penalty = gr.Slider(minimum=0.0, maximum=2.0, value=1.1, step=0.1, label="Repetition penalty")
|
170 |
|
171 |
-
blockchain_btn.click(lambda: select_topic("blockchain"), inputs=None, outputs=[
|
172 |
-
metaverse_btn.click(lambda: select_topic("metaverse"), inputs=None, outputs=[
|
173 |
-
payment_btn.click(lambda: select_topic("payment"), inputs=None, outputs=[
|
174 |
|
175 |
submit_btn.click(
|
176 |
respond,
|
177 |
-
inputs=[msg, chatbot, model, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty,
|
178 |
outputs=chatbot
|
179 |
)
|
180 |
|
181 |
msg.submit(
|
182 |
respond,
|
183 |
-
inputs=[msg, chatbot, model, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty,
|
184 |
outputs=chatbot
|
185 |
)
|
186 |
|
|
|
24 |
|
25 |
cv2.setNumThreads(1)
|
26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
@spaces.GPU(duration=20)
|
28 |
def respond(
|
29 |
message,
|
|
|
35 |
top_p,
|
36 |
top_k,
|
37 |
repeat_penalty,
|
38 |
+
selected_topic
|
39 |
):
|
40 |
chat_template = MessagesFormatterType.GEMMA_2
|
41 |
|
42 |
+
print("HISTORY SO FAR ", history)
|
43 |
+
print("Selected topic:", selected_topic)
|
44 |
|
45 |
+
if selected_topic:
|
46 |
+
query_engine = backend.create_index_for_query_engine(documents_paths[selected_topic])
|
47 |
+
full_prompt = backend.generate_prompt(query_engine, message)
|
48 |
+
gr.Info(f"Relevant context indexed from {selected_topic} docs...")
|
49 |
+
else:
|
50 |
+
query_engine = backend.load_index_for_query_engine()
|
51 |
+
full_prompt = backend.generate_prompt(query_engine, message)
|
52 |
+
gr.Info("Relevant context extracted from db...")
|
53 |
|
54 |
# Load model only if it's not already loaded or if a new model is selected
|
55 |
if backend.llm is None or backend.llm_model != model:
|
|
|
94 |
outputs = ""
|
95 |
for output in stream:
|
96 |
outputs += output
|
97 |
+
yield history + [[message, outputs]] # Use original message, not full_prompt
|
98 |
except Exception as e:
|
99 |
yield history + [[message, f"Error during response generation: {str(e)}"]]
|
100 |
|
101 |
def select_topic(topic):
|
102 |
+
return gr.update(visible=True), topic, gr.update(interactive=False), gr.update(interactive=False), gr.update(interactive=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
|
104 |
with gr.Blocks(css="""
|
105 |
.gradio-container {
|
|
|
114 |
metaverse_btn = gr.Button("๐ Metaverse", scale=1)
|
115 |
payment_btn = gr.Button("๐ณ Payment", scale=1)
|
116 |
|
117 |
+
selected_topic = gr.State(value="")
|
118 |
|
119 |
chatbot = gr.Chatbot(
|
120 |
scale=1,
|
|
|
129 |
show_label=False,
|
130 |
placeholder="Inserisci il tuo messaggio...",
|
131 |
container=False,
|
|
|
132 |
)
|
133 |
+
submit_btn = gr.Button("Invia", scale=1)
|
134 |
|
135 |
with gr.Accordion("Advanced Options", open=False):
|
136 |
model = gr.Dropdown([
|
|
|
146 |
value="gemma-2-2b-it-Q6_K_L.gguf",
|
147 |
label="Model"
|
148 |
)
|
149 |
+
system_message = gr.Textbox(value=""" Inizia presentandoti come "Odi", un assistente ricercatore italiano sviluppato dagli Osservatori del Politecnico di Milano, specializzato in innovazione digitale. Rispondi alle domande utilizzando esclusivamente i dati forniti tramite RAG. Se non trovi informazioni pertinenti, informa l'utente che non hai la risposta e suggerisci di contattare i responsabili dell'osservatorio, estraendo i loro nomi dai dati disponibili. Quando fornisci risposte, cita sempre il report specifico da cui hai ottenuto le informazioni. Utilizza la cronologia della chat e il contesto fornito per garantire risposte accurate e pertinenti. Non rispondere a domande fuori dal tuo ambito di competenza.""", label="System message")
|
|
|
|
|
|
|
150 |
max_tokens = gr.Slider(minimum=1, maximum=4096, value=3048, step=1, label="Max tokens")
|
151 |
temperature = gr.Slider(minimum=0.1, maximum=4.0, value=1.2, step=0.1, label="Temperature")
|
152 |
top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p")
|
153 |
top_k = gr.Slider(minimum=0, maximum=100, value=30, step=1, label="Top-k")
|
154 |
repeat_penalty = gr.Slider(minimum=0.0, maximum=2.0, value=1.1, step=0.1, label="Repetition penalty")
|
155 |
|
156 |
+
blockchain_btn.click(lambda: select_topic("blockchain"), inputs=None, outputs=[chatbot, selected_topic, blockchain_btn, metaverse_btn, payment_btn])
|
157 |
+
metaverse_btn.click(lambda: select_topic("metaverse"), inputs=None, outputs=[chatbot, selected_topic, blockchain_btn, metaverse_btn, payment_btn])
|
158 |
+
payment_btn.click(lambda: select_topic("payment"), inputs=None, outputs=[chatbot, selected_topic, blockchain_btn, metaverse_btn, payment_btn])
|
159 |
|
160 |
submit_btn.click(
|
161 |
respond,
|
162 |
+
inputs=[msg, chatbot, model, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty, selected_topic],
|
163 |
outputs=chatbot
|
164 |
)
|
165 |
|
166 |
msg.submit(
|
167 |
respond,
|
168 |
+
inputs=[msg, chatbot, model, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty, selected_topic],
|
169 |
outputs=chatbot
|
170 |
)
|
171 |
|