Update app.py
Browse files
app.py
CHANGED
@@ -1,63 +1,567 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
-
|
3 |
-
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
)
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import time
|
2 |
+
import re
|
3 |
+
import json
|
4 |
+
import os
|
5 |
+
from datetime import datetime
|
6 |
+
|
7 |
import gradio as gr
|
8 |
+
import torch
|
9 |
+
|
10 |
+
import modules.shared as shared
|
11 |
+
from modules import chat, ui as ui_module
|
12 |
+
from modules.utils import gradio
|
13 |
+
from modules.text_generation import generate_reply_HF, generate_reply_custom
|
14 |
+
from .llm_web_search import get_webpage_content, langchain_search_duckduckgo, langchain_search_searxng, Generator
|
15 |
+
from .langchain_websearch import LangchainCompressor
|
16 |
+
|
17 |
+
|
18 |
+
params = {
|
19 |
+
"display_name": "LLM Web Search",
|
20 |
+
"is_tab": True,
|
21 |
+
"enable": True,
|
22 |
+
"search results per query": 5,
|
23 |
+
"langchain similarity score threshold": 0.5,
|
24 |
+
"instant answers": True,
|
25 |
+
"regular search results": True,
|
26 |
+
"search command regex": "",
|
27 |
+
"default search command regex": r"Search_web\(\"(.*)\"\)",
|
28 |
+
"open url command regex": "",
|
29 |
+
"default open url command regex": r"Open_url\(\"(.*)\"\)",
|
30 |
+
"display search results in chat": True,
|
31 |
+
"display extracted URL content in chat": True,
|
32 |
+
"searxng url": "",
|
33 |
+
"cpu only": True,
|
34 |
+
"chunk size": 500,
|
35 |
+
"duckduckgo results per query": 10,
|
36 |
+
"append current datetime": False,
|
37 |
+
"default system prompt filename": None,
|
38 |
+
"force search prefix": "Search_web",
|
39 |
+
"ensemble weighting": 0.5,
|
40 |
+
"keyword retriever": "bm25",
|
41 |
+
"splade batch size": 2,
|
42 |
+
"chunking method": "character-based",
|
43 |
+
"chunker breakpoint_threshold_amount": 30
|
44 |
+
}
|
45 |
+
custom_system_message_filename = None
|
46 |
+
extension_path = os.path.dirname(os.path.abspath(__file__))
|
47 |
+
langchain_compressor = None
|
48 |
+
update_history = None
|
49 |
+
force_search = False
|
50 |
+
|
51 |
+
|
52 |
+
def setup():
|
53 |
+
"""
|
54 |
+
Is executed when the extension gets imported.
|
55 |
+
:return:
|
56 |
+
"""
|
57 |
+
global params
|
58 |
+
os.environ["TOKENIZERS_PARALLELISM"] = "true"
|
59 |
+
os.environ["QDRANT__TELEMETRY_DISABLED"] = "true"
|
60 |
+
|
61 |
+
try:
|
62 |
+
with open(os.path.join(extension_path, "settings.json"), "r") as f:
|
63 |
+
saved_params = json.load(f)
|
64 |
+
params.update(saved_params)
|
65 |
+
save_settings() # add keys of newly added feature to settings.json
|
66 |
+
except FileNotFoundError:
|
67 |
+
save_settings()
|
68 |
+
|
69 |
+
if not os.path.exists(os.path.join(extension_path, "system_prompts")):
|
70 |
+
os.makedirs(os.path.join(extension_path, "system_prompts"))
|
71 |
+
|
72 |
+
toggle_extension(params["enable"])
|
73 |
+
|
74 |
+
|
75 |
+
def save_settings():
|
76 |
+
global params
|
77 |
+
with open(os.path.join(extension_path, "settings.json"), "w") as f:
|
78 |
+
json.dump(params, f, indent=4)
|
79 |
+
current_datetime = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
|
80 |
+
return gr.HTML(f'<span style="color:lawngreen"> Settings were saved at {current_datetime}</span>',
|
81 |
+
visible=True)
|
82 |
+
|
83 |
+
|
84 |
+
def toggle_extension(_enable: bool):
|
85 |
+
global langchain_compressor, custom_system_message_filename
|
86 |
+
if _enable:
|
87 |
+
langchain_compressor = LangchainCompressor(device="cpu" if params["cpu only"] else "cuda",
|
88 |
+
keyword_retriever=params["keyword retriever"],
|
89 |
+
model_cache_dir=os.path.join(extension_path, "hf_models"))
|
90 |
+
compressor_model = langchain_compressor.embeddings.client
|
91 |
+
compressor_model.to(compressor_model._target_device)
|
92 |
+
custom_system_message_filename = params.get("default system prompt filename")
|
93 |
+
else:
|
94 |
+
if not params["cpu only"] and 'langchain_compressor' in globals(): # free some VRAM
|
95 |
+
model_attrs = ["embeddings", "splade_doc_model", "splade_query_model"]
|
96 |
+
for model_attr in model_attrs:
|
97 |
+
if hasattr(langchain_compressor, model_attr):
|
98 |
+
model = getattr(langchain_compressor, model_attr)
|
99 |
+
if hasattr(model, "client"):
|
100 |
+
model.client.to("cpu")
|
101 |
+
del model.client
|
102 |
+
else:
|
103 |
+
if hasattr(model, "to"):
|
104 |
+
model.to("cpu")
|
105 |
+
del model
|
106 |
+
torch.cuda.empty_cache()
|
107 |
+
params.update({"enable": _enable})
|
108 |
+
return _enable
|
109 |
+
|
110 |
+
|
111 |
+
def get_available_system_prompts():
|
112 |
+
try:
|
113 |
+
return ["None"] + sorted(os.listdir(os.path.join(extension_path, "system_prompts")))
|
114 |
+
except FileNotFoundError:
|
115 |
+
return ["None"]
|
116 |
+
|
117 |
+
|
118 |
+
def load_system_prompt(filename: str or None):
|
119 |
+
global custom_system_message_filename
|
120 |
+
if not filename:
|
121 |
+
return
|
122 |
+
if filename == "None" or filename == "Select custom system message to load...":
|
123 |
+
custom_system_message_filename = None
|
124 |
+
return ""
|
125 |
+
with open(os.path.join(extension_path, "system_prompts", filename), "r") as f:
|
126 |
+
prompt_str = f.read()
|
127 |
+
|
128 |
+
if params["append current datetime"]:
|
129 |
+
prompt_str += f"\nDate and time of conversation: {datetime.now().strftime('%A %d %B %Y %H:%M')}"
|
130 |
+
|
131 |
+
shared.settings['custom_system_message'] = prompt_str
|
132 |
+
custom_system_message_filename = filename
|
133 |
+
return prompt_str
|
134 |
+
|
135 |
+
|
136 |
+
def save_system_prompt(filename, prompt):
|
137 |
+
if not filename:
|
138 |
+
return
|
139 |
+
|
140 |
+
with open(os.path.join(extension_path, "system_prompts", filename), "w") as f:
|
141 |
+
f.write(prompt)
|
142 |
+
|
143 |
+
return gr.HTML(f'<span style="color:lawngreen"> Saved successfully</span>',
|
144 |
+
visible=True)
|
145 |
+
|
146 |
+
|
147 |
+
def check_file_exists(filename):
|
148 |
+
if filename == "":
|
149 |
+
return gr.HTML("", visible=False)
|
150 |
+
if os.path.exists(os.path.join(extension_path, "system_prompts", filename)):
|
151 |
+
return gr.HTML(f'<span style="color:orange"> Warning: Filename already exists</span>', visible=True)
|
152 |
+
return gr.HTML("", visible=False)
|
153 |
+
|
154 |
+
|
155 |
+
def timeout_save_message():
|
156 |
+
time.sleep(2)
|
157 |
+
return gr.HTML("", visible=False)
|
158 |
+
|
159 |
+
|
160 |
+
def deactivate_system_prompt():
|
161 |
+
shared.settings['custom_system_message'] = None
|
162 |
+
return "None"
|
163 |
+
|
164 |
+
|
165 |
+
def toggle_forced_search(value):
|
166 |
+
global force_search
|
167 |
+
force_search = value
|
168 |
+
|
169 |
+
|
170 |
+
def ui():
|
171 |
+
"""
|
172 |
+
Creates custom gradio elements when the UI is launched.
|
173 |
+
:return:
|
174 |
+
"""
|
175 |
+
# Inject custom system message into the main textbox if a default one is set
|
176 |
+
shared.gradio['custom_system_message'].value = load_system_prompt(custom_system_message_filename)
|
177 |
+
|
178 |
+
def update_result_type_setting(choice: str):
|
179 |
+
if choice == "Instant answers":
|
180 |
+
params.update({"instant answers": True})
|
181 |
+
params.update({"regular search results": False})
|
182 |
+
elif choice == "Regular results":
|
183 |
+
params.update({"instant answers": False})
|
184 |
+
params.update({"regular search results": True})
|
185 |
+
elif choice == "Regular results and instant answers":
|
186 |
+
params.update({"instant answers": True})
|
187 |
+
params.update({"regular search results": True})
|
188 |
+
|
189 |
+
def update_regex_setting(input_str: str, setting_key: str, error_html_element: gr.component):
|
190 |
+
if input_str == "":
|
191 |
+
params.update({setting_key: params[f"default {setting_key}"]})
|
192 |
+
return {error_html_element: gr.HTML("", visible=False)}
|
193 |
+
try:
|
194 |
+
compiled = re.compile(input_str)
|
195 |
+
if compiled.groups > 1:
|
196 |
+
raise re.error(f"Only 1 capturing group allowed in regex, but there are {compiled.groups}.")
|
197 |
+
params.update({setting_key: input_str})
|
198 |
+
return {error_html_element: gr.HTML("", visible=False)}
|
199 |
+
except re.error as e:
|
200 |
+
return {error_html_element: gr.HTML(f'<span style="color:red"> Invalid regex. {str(e).capitalize()}</span>',
|
201 |
+
visible=True)}
|
202 |
+
|
203 |
+
def update_default_custom_system_message(check: bool):
|
204 |
+
if check:
|
205 |
+
params.update({"default system prompt filename": custom_system_message_filename})
|
206 |
+
else:
|
207 |
+
params.update({"default system prompt filename": None})
|
208 |
+
|
209 |
+
with gr.Row():
|
210 |
+
enable = gr.Checkbox(value=lambda: params['enable'], label='Enable LLM web search')
|
211 |
+
use_cpu_only = gr.Checkbox(value=lambda: params['cpu only'],
|
212 |
+
label='Run extension on CPU only '
|
213 |
+
'(Save settings and restart for the change to take effect)')
|
214 |
+
with gr.Column():
|
215 |
+
save_settings_btn = gr.Button("Save settings")
|
216 |
+
saved_success_elem = gr.HTML("", visible=False)
|
217 |
+
|
218 |
+
with gr.Row():
|
219 |
+
result_radio = gr.Radio(
|
220 |
+
["Regular results", "Regular results and instant answers"],
|
221 |
+
label="What kind of search results should be returned?",
|
222 |
+
value=lambda: "Regular results and instant answers" if
|
223 |
+
(params["regular search results"] and params["instant answers"]) else "Regular results"
|
224 |
+
)
|
225 |
+
with gr.Column():
|
226 |
+
search_command_regex = gr.Textbox(label="Search command regex string",
|
227 |
+
placeholder=params["default search command regex"],
|
228 |
+
value=lambda: params["search command regex"])
|
229 |
+
search_command_regex_error_label = gr.HTML("", visible=False)
|
230 |
+
|
231 |
+
with gr.Column():
|
232 |
+
open_url_command_regex = gr.Textbox(label="Open URL command regex string",
|
233 |
+
placeholder=params["default open url command regex"],
|
234 |
+
value=lambda: params["open url command regex"])
|
235 |
+
open_url_command_regex_error_label = gr.HTML("", visible=False)
|
236 |
+
|
237 |
+
with gr.Column():
|
238 |
+
show_results = gr.Checkbox(value=lambda: params['display search results in chat'],
|
239 |
+
label='Display search results in chat')
|
240 |
+
show_url_content = gr.Checkbox(value=lambda: params['display extracted URL content in chat'],
|
241 |
+
label='Display extracted URL content in chat')
|
242 |
+
gr.Markdown(value='---')
|
243 |
+
with gr.Row():
|
244 |
+
with gr.Column():
|
245 |
+
gr.Markdown(value='#### Load custom system message\n'
|
246 |
+
'Select a saved custom system message from within the system_prompts folder or "None" '
|
247 |
+
'to clear the selection')
|
248 |
+
system_prompt = gr.Dropdown(
|
249 |
+
choices=get_available_system_prompts(), label="Select custom system message",
|
250 |
+
value=lambda: 'Select custom system message to load...' if custom_system_message_filename is None else
|
251 |
+
custom_system_message_filename, elem_classes='slim-dropdown')
|
252 |
+
with gr.Row():
|
253 |
+
set_system_message_as_default = gr.Checkbox(
|
254 |
+
value=lambda: custom_system_message_filename == params["default system prompt filename"],
|
255 |
+
label='Set this custom system message as the default')
|
256 |
+
refresh_button = ui_module.create_refresh_button(system_prompt, lambda: None,
|
257 |
+
lambda: {'choices': get_available_system_prompts()},
|
258 |
+
'refresh-button', interactive=True)
|
259 |
+
refresh_button.elem_id = "custom-sysprompt-refresh"
|
260 |
+
delete_button = gr.Button('🗑️', elem_classes='refresh-button', interactive=True)
|
261 |
+
append_datetime = gr.Checkbox(value=lambda: params['append current datetime'],
|
262 |
+
label='Append current date and time when loading custom system message')
|
263 |
+
with gr.Column():
|
264 |
+
gr.Markdown(value='#### Create custom system message')
|
265 |
+
system_prompt_text = gr.Textbox(label="Custom system message", lines=3,
|
266 |
+
value=lambda: load_system_prompt(custom_system_message_filename))
|
267 |
+
sys_prompt_filename = gr.Text(label="Filename")
|
268 |
+
sys_prompt_save_button = gr.Button("Save Custom system message")
|
269 |
+
system_prompt_saved_success_elem = gr.HTML("", visible=False)
|
270 |
+
|
271 |
+
gr.Markdown(value='---')
|
272 |
+
with gr.Accordion("Advanced settings", open=False):
|
273 |
+
ensemble_weighting = gr.Slider(minimum=0, maximum=1, step=0.05, value=lambda: params["ensemble weighting"],
|
274 |
+
label="Ensemble Weighting", info="Smaller values = More keyword oriented, "
|
275 |
+
"Larger values = More focus on semantic similarity")
|
276 |
+
with gr.Row():
|
277 |
+
keyword_retriever = gr.Radio([("Okapi BM25", "bm25"),("SPLADE", "splade")], label="Sparse keyword retriever",
|
278 |
+
info="For change to take effect, toggle the extension off and on again",
|
279 |
+
value=lambda: params["keyword retriever"])
|
280 |
+
splade_batch_size = gr.Slider(minimum=2, maximum=256, step=2, value=lambda: params["splade batch size"],
|
281 |
+
label="SPLADE batch size",
|
282 |
+
info="Smaller values = Slower retrieval (but lower VRAM usage), "
|
283 |
+
"Larger values = Faster retrieval (but higher VRAM usage). "
|
284 |
+
"A good trade-off seems to be setting it = 8",
|
285 |
+
precision=0)
|
286 |
+
with gr.Row():
|
287 |
+
chunker = gr.Radio([("Character-based", "character-based"),
|
288 |
+
("Semantic", "semantic")], label="Chunking method",
|
289 |
+
value=lambda: params["chunking method"])
|
290 |
+
chunker_breakpoint_threshold_amount = gr.Slider(minimum=1, maximum=100, step=1,
|
291 |
+
value=lambda: params["chunker breakpoint_threshold_amount"],
|
292 |
+
label="Semantic chunking: sentence split threshold (%)",
|
293 |
+
info="Defines how different two consecutive sentences have"
|
294 |
+
" to be for them to be split into separate chunks",
|
295 |
+
precision=0)
|
296 |
+
gr.Markdown("**Note: Changing the following might result in DuckDuckGo rate limiting or the LM being overwhelmed**")
|
297 |
+
num_search_results = gr.Number(label="Max. search results to return per query", minimum=1, maximum=100,
|
298 |
+
value=lambda: params["search results per query"], precision=0)
|
299 |
+
num_process_search_results = gr.Number(label="Number of search results to process per query", minimum=1,
|
300 |
+
maximum=100, value=lambda: params["duckduckgo results per query"],
|
301 |
+
precision=0)
|
302 |
+
langchain_similarity_threshold = gr.Number(label="Langchain Similarity Score Threshold", minimum=0., maximum=1.,
|
303 |
+
value=lambda: params["langchain similarity score threshold"])
|
304 |
+
chunk_size = gr.Number(label="Max. chunk size", info="The maximal size of the individual chunks that each webpage will"
|
305 |
+
" be split into, in characters", minimum=2, maximum=10000,
|
306 |
+
value=lambda: params["chunk size"], precision=0)
|
307 |
+
|
308 |
+
with gr.Row():
|
309 |
+
searxng_url = gr.Textbox(label="SearXNG URL",
|
310 |
+
value=lambda: params["searxng url"])
|
311 |
+
|
312 |
+
# Event functions to update the parameters in the backend
|
313 |
+
enable.input(toggle_extension, enable, enable)
|
314 |
+
use_cpu_only.change(lambda x: params.update({"cpu only": x}), use_cpu_only, None)
|
315 |
+
save_settings_btn.click(save_settings, None, [saved_success_elem])
|
316 |
+
ensemble_weighting.change(lambda x: params.update({"ensemble weighting": x}), ensemble_weighting, None)
|
317 |
+
keyword_retriever.change(lambda x: params.update({"keyword retriever": x}), keyword_retriever, None)
|
318 |
+
splade_batch_size.change(lambda x: params.update({"splade batch size": x}), splade_batch_size, None)
|
319 |
+
chunker.change(lambda x: params.update({"chunking method": x}), chunker, None)
|
320 |
+
chunker_breakpoint_threshold_amount.change(lambda x: params.update({"chunker breakpoint_threshold_amount": x}),
|
321 |
+
chunker_breakpoint_threshold_amount, None)
|
322 |
+
num_search_results.change(lambda x: params.update({"search results per query": x}), num_search_results, None)
|
323 |
+
num_process_search_results.change(lambda x: params.update({"duckduckgo results per query": x}),
|
324 |
+
num_process_search_results, None)
|
325 |
+
langchain_similarity_threshold.change(lambda x: params.update({"langchain similarity score threshold": x}),
|
326 |
+
langchain_similarity_threshold, None)
|
327 |
+
chunk_size.change(lambda x: params.update({"chunk size": x}), chunk_size, None)
|
328 |
+
result_radio.change(update_result_type_setting, result_radio, None)
|
329 |
+
|
330 |
+
search_command_regex.change(lambda x: update_regex_setting(x, "search command regex",
|
331 |
+
search_command_regex_error_label),
|
332 |
+
search_command_regex, search_command_regex_error_label, show_progress="hidden")
|
333 |
+
|
334 |
+
open_url_command_regex.change(lambda x: update_regex_setting(x, "open url command regex",
|
335 |
+
open_url_command_regex_error_label),
|
336 |
+
open_url_command_regex, open_url_command_regex_error_label, show_progress="hidden")
|
337 |
+
|
338 |
+
show_results.change(lambda x: params.update({"display search results in chat": x}), show_results, None)
|
339 |
+
show_url_content.change(lambda x: params.update({"display extracted URL content in chat": x}), show_url_content,
|
340 |
+
None)
|
341 |
+
searxng_url.change(lambda x: params.update({"searxng url": x}), searxng_url, None)
|
342 |
+
|
343 |
+
delete_button.click(
|
344 |
+
lambda x: x, system_prompt, gradio('delete_filename')).then(
|
345 |
+
lambda: os.path.join(extension_path, "system_prompts", ""), None, gradio('delete_root')).then(
|
346 |
+
lambda: gr.update(visible=True), None, gradio('file_deleter'))
|
347 |
+
shared.gradio['delete_confirm'].click(
|
348 |
+
lambda: "None", None, system_prompt).then(
|
349 |
+
None, None, None, _js="() => { document.getElementById('custom-sysprompt-refresh').click() }")
|
350 |
+
system_prompt.change(load_system_prompt, system_prompt, shared.gradio['custom_system_message'])
|
351 |
+
system_prompt.change(load_system_prompt, system_prompt, system_prompt_text)
|
352 |
+
# restore checked state if chosen system prompt matches the default
|
353 |
+
system_prompt.change(lambda x: x == params["default system prompt filename"], system_prompt,
|
354 |
+
set_system_message_as_default)
|
355 |
+
sys_prompt_filename.change(check_file_exists, sys_prompt_filename, system_prompt_saved_success_elem)
|
356 |
+
sys_prompt_save_button.click(save_system_prompt, [sys_prompt_filename, system_prompt_text],
|
357 |
+
system_prompt_saved_success_elem,
|
358 |
+
show_progress="hidden").then(timeout_save_message,
|
359 |
+
None,
|
360 |
+
system_prompt_saved_success_elem,
|
361 |
+
_js="() => { document.getElementById('custom-sysprompt-refresh').click() }",
|
362 |
+
show_progress="hidden").then(lambda: "", None,
|
363 |
+
sys_prompt_filename,
|
364 |
+
show_progress="hidden")
|
365 |
+
append_datetime.change(lambda x: params.update({"append current datetime": x}), append_datetime, None)
|
366 |
+
# '.input' = only triggers when user changes the value of the component, not a function
|
367 |
+
set_system_message_as_default.input(update_default_custom_system_message, set_system_message_as_default, None)
|
368 |
+
|
369 |
+
# A dummy checkbox to enable the actual "Force web search" checkbox to trigger a gradio event
|
370 |
+
force_search_checkbox = gr.Checkbox(value=False, visible=False, elem_id="Force-search-checkbox")
|
371 |
+
force_search_checkbox.change(toggle_forced_search, force_search_checkbox, None)
|
372 |
+
|
373 |
+
|
374 |
+
def custom_generate_reply(question, original_question, seed, state, stopping_strings, is_chat):
|
375 |
+
"""
|
376 |
+
Overrides the main text generation function.
|
377 |
+
:return:
|
378 |
+
"""
|
379 |
+
global update_history, langchain_compressor
|
380 |
+
if shared.model.__class__.__name__ in ['LlamaCppModel', 'RWKVModel', 'ExllamaModel', 'Exllamav2Model',
|
381 |
+
'CtransformersModel']:
|
382 |
+
generate_func = generate_reply_custom
|
383 |
+
else:
|
384 |
+
generate_func = generate_reply_HF
|
385 |
+
|
386 |
+
if not params['enable']:
|
387 |
+
for reply in generate_func(question, original_question, seed, state, stopping_strings, is_chat=is_chat):
|
388 |
+
yield reply
|
389 |
+
return
|
390 |
+
|
391 |
+
web_search = False
|
392 |
+
read_webpage = False
|
393 |
+
max_search_results = int(params["search results per query"])
|
394 |
+
instant_answers = params["instant answers"]
|
395 |
+
# regular_search_results = params["regular search results"]
|
396 |
+
|
397 |
+
langchain_compressor.num_results = int(params["duckduckgo results per query"])
|
398 |
+
langchain_compressor.similarity_threshold = params["langchain similarity score threshold"]
|
399 |
+
langchain_compressor.chunk_size = params["chunk size"]
|
400 |
+
langchain_compressor.ensemble_weighting = params["ensemble weighting"]
|
401 |
+
langchain_compressor.splade_batch_size = params["splade batch size"]
|
402 |
+
langchain_compressor.chunking_method = params["chunking method"]
|
403 |
+
langchain_compressor.chunker_breakpoint_threshold_amount = params["chunker breakpoint_threshold_amount"]
|
404 |
+
|
405 |
+
search_command_regex = params["search command regex"]
|
406 |
+
open_url_command_regex = params["open url command regex"]
|
407 |
+
searxng_url = params["searxng url"]
|
408 |
+
display_search_results = params["display search results in chat"]
|
409 |
+
display_webpage_content = params["display extracted URL content in chat"]
|
410 |
+
|
411 |
+
if search_command_regex == "":
|
412 |
+
search_command_regex = params["default search command regex"]
|
413 |
+
if open_url_command_regex == "":
|
414 |
+
open_url_command_regex = params["default open url command regex"]
|
415 |
+
|
416 |
+
compiled_search_command_regex = re.compile(search_command_regex)
|
417 |
+
compiled_open_url_command_regex = re.compile(open_url_command_regex)
|
418 |
+
|
419 |
+
if force_search:
|
420 |
+
question += f" {params['force search prefix']}"
|
421 |
+
|
422 |
+
reply = None
|
423 |
+
for reply in generate_func(question, original_question, seed, state, stopping_strings, is_chat=is_chat):
|
424 |
+
|
425 |
+
if force_search:
|
426 |
+
reply = params["force search prefix"] + reply
|
427 |
+
|
428 |
+
search_re_match = compiled_search_command_regex.search(reply)
|
429 |
+
if search_re_match is not None:
|
430 |
+
yield reply
|
431 |
+
original_model_reply = reply
|
432 |
+
web_search = True
|
433 |
+
search_term = search_re_match.group(1)
|
434 |
+
print(f"LLM_Web_search | Searching for {search_term}...")
|
435 |
+
reply += "\n```plaintext"
|
436 |
+
reply += "\nSearch tool:\n"
|
437 |
+
if searxng_url == "":
|
438 |
+
search_generator = Generator(langchain_search_duckduckgo(search_term,
|
439 |
+
langchain_compressor,
|
440 |
+
max_search_results,
|
441 |
+
instant_answers))
|
442 |
+
else:
|
443 |
+
search_generator = Generator(langchain_search_searxng(search_term,
|
444 |
+
searxng_url,
|
445 |
+
langchain_compressor,
|
446 |
+
max_search_results))
|
447 |
+
try:
|
448 |
+
for status_message in search_generator:
|
449 |
+
yield original_model_reply + f"\n*{status_message}*"
|
450 |
+
search_results = search_generator.value
|
451 |
+
except Exception as exc:
|
452 |
+
exception_message = str(exc)
|
453 |
+
reply += f"The search tool encountered an error: {exception_message}"
|
454 |
+
print(f'LLM_Web_search | {search_term} generated an exception: {exception_message}')
|
455 |
+
else:
|
456 |
+
if search_results != "":
|
457 |
+
reply += search_results
|
458 |
+
else:
|
459 |
+
reply += f"\nThe search tool did not return any results."
|
460 |
+
reply += "```"
|
461 |
+
if display_search_results:
|
462 |
+
yield reply
|
463 |
+
break
|
464 |
+
|
465 |
+
open_url_re_match = compiled_open_url_command_regex.search(reply)
|
466 |
+
if open_url_re_match is not None:
|
467 |
+
yield reply
|
468 |
+
original_model_reply = reply
|
469 |
+
read_webpage = True
|
470 |
+
url = open_url_re_match.group(1)
|
471 |
+
print(f"LLM_Web_search | Reading {url}...")
|
472 |
+
reply += "\n```plaintext"
|
473 |
+
reply += "\nURL opener tool:\n"
|
474 |
+
try:
|
475 |
+
webpage_content = get_webpage_content(url)
|
476 |
+
except Exception as exc:
|
477 |
+
reply += f"Couldn't open {url}. Error message: {str(exc)}"
|
478 |
+
print(f'LLM_Web_search | {url} generated an exception: {str(exc)}')
|
479 |
+
else:
|
480 |
+
reply += f"\nText content of {url}:\n"
|
481 |
+
reply += webpage_content
|
482 |
+
reply += "```\n"
|
483 |
+
if display_webpage_content:
|
484 |
+
yield reply
|
485 |
+
break
|
486 |
+
yield reply
|
487 |
+
|
488 |
+
if web_search or read_webpage:
|
489 |
+
display_results = web_search and display_search_results or read_webpage and display_webpage_content
|
490 |
+
# Add results to context and continue model output
|
491 |
+
new_question = chat.generate_chat_prompt(f"{question}{reply}", state)
|
492 |
+
new_reply = ""
|
493 |
+
for new_reply in generate_func(new_question, new_question, seed, state,
|
494 |
+
stopping_strings, is_chat=is_chat):
|
495 |
+
if display_results:
|
496 |
+
yield f"{reply}\n{new_reply}"
|
497 |
+
else:
|
498 |
+
yield f"{original_model_reply}\n{new_reply}"
|
499 |
+
|
500 |
+
if not display_results:
|
501 |
+
update_history = [state["textbox"], f"{reply}\n{new_reply}"]
|
502 |
+
|
503 |
+
|
504 |
+
def output_modifier(string, state, is_chat=False):
|
505 |
+
"""
|
506 |
+
Modifies the output string before it is presented in the UI. In chat mode,
|
507 |
+
it is applied to the bot's reply. Otherwise, it is applied to the entire
|
508 |
+
output.
|
509 |
+
:param string:
|
510 |
+
:param state:
|
511 |
+
:param is_chat:
|
512 |
+
:return:
|
513 |
+
"""
|
514 |
+
return string
|
515 |
+
|
516 |
+
|
517 |
+
def custom_css():
|
518 |
+
"""
|
519 |
+
Returns custom CSS as a string. It is applied whenever the web UI is loaded.
|
520 |
+
:return:
|
521 |
+
"""
|
522 |
+
return ''
|
523 |
+
|
524 |
+
|
525 |
+
def custom_js():
|
526 |
+
"""
|
527 |
+
Returns custom javascript as a string. It is applied whenever the web UI is
|
528 |
+
loaded.
|
529 |
+
:return:
|
530 |
+
"""
|
531 |
+
with open(os.path.join(extension_path, "script.js"), "r") as f:
|
532 |
+
return f.read()
|
533 |
+
|
534 |
+
|
535 |
+
def chat_input_modifier(text, visible_text, state):
|
536 |
+
"""
|
537 |
+
Modifies both the visible and internal inputs in chat mode. Can be used to
|
538 |
+
hijack the chat input with custom content.
|
539 |
+
:param text:
|
540 |
+
:param visible_text:
|
541 |
+
:param state:
|
542 |
+
:return:
|
543 |
+
"""
|
544 |
+
return text, visible_text
|
545 |
+
|
546 |
+
|
547 |
+
def state_modifier(state):
|
548 |
+
"""
|
549 |
+
Modifies the dictionary containing the UI input parameters before it is
|
550 |
+
used by the text generation functions.
|
551 |
+
:param state:
|
552 |
+
:return:
|
553 |
+
"""
|
554 |
+
return state
|
555 |
+
|
556 |
+
|
557 |
+
def history_modifier(history):
|
558 |
+
"""
|
559 |
+
Modifies the chat history before the text generation in chat mode begins.
|
560 |
+
:param history:
|
561 |
+
:return:
|
562 |
+
"""
|
563 |
+
global update_history
|
564 |
+
if update_history:
|
565 |
+
history["internal"].append(update_history)
|
566 |
+
update_history = None
|
567 |
+
return history
|