Update src/txagent/txagent.py
Browse files- src/txagent/txagent.py +322 -511
src/txagent/txagent.py
CHANGED
@@ -11,19 +11,23 @@ import types
|
|
11 |
from tooluniverse import ToolUniverse
|
12 |
from gradio import ChatMessage
|
13 |
from .toolrag import ToolRAGModel
|
|
|
|
|
14 |
|
15 |
-
|
|
|
16 |
|
|
|
17 |
|
18 |
class TxAgent:
|
19 |
def __init__(self, model_name,
|
20 |
rag_model_name,
|
21 |
-
tool_files_dict=None,
|
22 |
enable_finish=True,
|
23 |
-
enable_rag=
|
24 |
enable_summary=False,
|
25 |
init_rag_num=0,
|
26 |
-
step_rag_num=
|
27 |
summary_mode='step',
|
28 |
summary_skip_last_k=0,
|
29 |
summary_context_length=None,
|
@@ -32,8 +36,7 @@ class TxAgent:
|
|
32 |
seed=None,
|
33 |
enable_checker=False,
|
34 |
enable_chat=False,
|
35 |
-
additional_default_tools=None
|
36 |
-
):
|
37 |
self.model_name = model_name
|
38 |
self.tokenizer = None
|
39 |
self.terminators = None
|
@@ -42,10 +45,9 @@ class TxAgent:
|
|
42 |
self.model = None
|
43 |
self.rag_model = ToolRAGModel(rag_model_name)
|
44 |
self.tooluniverse = None
|
45 |
-
|
46 |
-
self.prompt_multi_step = "You are a helpful assistant that will solve problems through detailed, step-by-step reasoning and actions based on your reasoning. Typically, your actions will use the provided functions. You have access to the following functions."
|
47 |
self.self_prompt = "Strictly follow the instruction."
|
48 |
-
self.chat_prompt = "You are helpful assistant
|
49 |
self.enable_finish = enable_finish
|
50 |
self.enable_rag = enable_rag
|
51 |
self.enable_summary = enable_summary
|
@@ -59,16 +61,11 @@ class TxAgent:
|
|
59 |
self.seed = seed
|
60 |
self.enable_checker = enable_checker
|
61 |
self.additional_default_tools = additional_default_tools
|
62 |
-
|
63 |
|
64 |
def init_model(self):
|
65 |
self.load_models()
|
66 |
self.load_tooluniverse()
|
67 |
-
self.load_tool_desc_embedding()
|
68 |
-
|
69 |
-
def print_self_values(self):
|
70 |
-
for attr, value in self.__dict__.items():
|
71 |
-
print(f"{attr}: {value}")
|
72 |
|
73 |
def load_models(self, model_name=None):
|
74 |
if model_name is not None:
|
@@ -76,10 +73,10 @@ class TxAgent:
|
|
76 |
return f"The model {model_name} is already loaded."
|
77 |
self.model_name = model_name
|
78 |
|
79 |
-
self.model = LLM(model=self.model_name)
|
80 |
self.chat_template = Template(self.model.get_tokenizer().chat_template)
|
81 |
self.tokenizer = self.model.get_tokenizer()
|
82 |
-
|
83 |
return f"Model {model_name} loaded successfully."
|
84 |
|
85 |
def load_tooluniverse(self):
|
@@ -88,9 +85,16 @@ class TxAgent:
|
|
88 |
special_tools = self.tooluniverse.prepare_tool_prompts(
|
89 |
self.tooluniverse.tool_category_dicts["special_tools"])
|
90 |
self.special_tools_name = [tool['name'] for tool in special_tools]
|
|
|
91 |
|
92 |
def load_tool_desc_embedding(self):
|
93 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
94 |
|
95 |
def rag_infer(self, query, top_k=5):
|
96 |
return self.rag_model.rag_infer(query, top_k)
|
@@ -103,10 +107,6 @@ class TxAgent:
|
|
103 |
call_agent_level += 1
|
104 |
if call_agent_level >= 2:
|
105 |
call_agent = False
|
106 |
-
|
107 |
-
if not call_agent:
|
108 |
-
picked_tools_prompt += self.tool_RAG(
|
109 |
-
message=message, rag_num=self.init_rag_num)
|
110 |
return picked_tools_prompt, call_agent_level
|
111 |
|
112 |
def initialize_conversation(self, message, conversation=None, history=None):
|
@@ -115,85 +115,56 @@ class TxAgent:
|
|
115 |
|
116 |
conversation = self.set_system_prompt(
|
117 |
conversation, self.prompt_multi_step)
|
118 |
-
if history
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
if history[i]['role'] == 'user':
|
125 |
-
if i-1 >= 0 and history[i-1]['role'] == 'assistant':
|
126 |
-
conversation.append(
|
127 |
-
{"role": "assistant", "content": history[i-1]['content']})
|
128 |
-
conversation.append(
|
129 |
-
{"role": "user", "content": history[i]['content']})
|
130 |
-
if i == len(history)-1 and history[i]['role'] == 'assistant':
|
131 |
-
conversation.append(
|
132 |
-
{"role": "assistant", "content": history[i]['content']})
|
133 |
-
|
134 |
conversation.append({"role": "user", "content": message})
|
135 |
-
|
136 |
return conversation
|
137 |
|
138 |
def tool_RAG(self, message=None,
|
139 |
picked_tool_names=None,
|
140 |
existing_tools_prompt=[],
|
141 |
-
rag_num=
|
142 |
return_call_result=False):
|
143 |
-
|
|
|
|
|
144 |
if picked_tool_names is None:
|
145 |
assert picked_tool_names is not None or message is not None
|
146 |
picked_tool_names = self.rag_infer(
|
147 |
-
message, top_k=rag_num*extra_factor)
|
148 |
|
149 |
-
picked_tool_names_no_special = []
|
150 |
-
for tool in picked_tool_names:
|
151 |
-
if tool not in self.special_tools_name:
|
152 |
-
picked_tool_names_no_special.append(tool)
|
153 |
-
picked_tool_names_no_special = picked_tool_names_no_special[:rag_num]
|
154 |
picked_tool_names = picked_tool_names_no_special[:rag_num]
|
155 |
|
156 |
picked_tools = self.tooluniverse.get_tool_by_name(picked_tool_names)
|
157 |
-
picked_tools_prompt = self.tooluniverse.prepare_tool_prompts(
|
158 |
-
|
159 |
if return_call_result:
|
160 |
return picked_tools_prompt, picked_tool_names
|
161 |
return picked_tools_prompt
|
162 |
|
163 |
def add_special_tools(self, tools, call_agent=False):
|
164 |
if self.enable_finish:
|
165 |
-
tools.append(self.tooluniverse.get_one_tool_by_one_name(
|
166 |
-
|
167 |
-
print("Finish tool is added")
|
168 |
if call_agent:
|
169 |
-
tools.append(self.tooluniverse.get_one_tool_by_one_name(
|
170 |
-
|
171 |
-
print("CallAgent tool is added")
|
172 |
-
else:
|
173 |
-
if self.enable_rag:
|
174 |
-
tools.append(self.tooluniverse.get_one_tool_by_one_name(
|
175 |
-
'Tool_RAG', return_prompt=True))
|
176 |
-
print("Tool_RAG tool is added")
|
177 |
-
|
178 |
-
if self.additional_default_tools is not None:
|
179 |
-
for each_tool_name in self.additional_default_tools:
|
180 |
-
tool_prompt = self.tooluniverse.get_one_tool_by_one_name(
|
181 |
-
each_tool_name, return_prompt=True)
|
182 |
-
if tool_prompt is not None:
|
183 |
-
print(f"{each_tool_name} tool is added")
|
184 |
-
tools.append(tool_prompt)
|
185 |
return tools
|
186 |
|
187 |
def add_finish_tools(self, tools):
|
188 |
-
tools.append(self.tooluniverse.get_one_tool_by_one_name(
|
189 |
-
|
190 |
-
print("Finish tool is added")
|
191 |
return tools
|
192 |
|
193 |
def set_system_prompt(self, conversation, sys_prompt):
|
194 |
-
if
|
195 |
-
conversation.append(
|
196 |
-
{"role": "system", "content": sys_prompt})
|
197 |
else:
|
198 |
conversation[0] = {"role": "system", "content": sys_prompt}
|
199 |
return conversation
|
@@ -205,25 +176,23 @@ class TxAgent:
|
|
205 |
call_agent=False,
|
206 |
call_agent_level=None,
|
207 |
temperature=None):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
208 |
|
209 |
-
function_call_json, message = self.tooluniverse.extract_function_call_json(
|
210 |
-
fcall_str, return_message=return_message, verbose=False)
|
211 |
call_results = []
|
212 |
special_tool_call = ''
|
213 |
-
if function_call_json
|
214 |
if isinstance(function_call_json, list):
|
215 |
for i in range(len(function_call_json)):
|
216 |
-
|
217 |
if function_call_json[i]["name"] == 'Finish':
|
218 |
special_tool_call = 'Finish'
|
219 |
break
|
220 |
-
elif function_call_json[i]["name"] == 'Tool_RAG':
|
221 |
-
new_tools_prompt, call_result = self.tool_RAG(
|
222 |
-
message=message,
|
223 |
-
existing_tools_prompt=existing_tools_prompt,
|
224 |
-
rag_num=self.step_rag_num,
|
225 |
-
return_call_result=True)
|
226 |
-
existing_tools_prompt += new_tools_prompt
|
227 |
elif function_call_json[i]["name"] == 'CallAgent':
|
228 |
if call_agent_level < 2 and call_agent:
|
229 |
solution_plan = function_call_json[i]['arguments']['solution']
|
@@ -234,27 +203,27 @@ class TxAgent:
|
|
234 |
)
|
235 |
call_result = self.run_multistep_agent(
|
236 |
full_message, temperature=temperature,
|
237 |
-
max_new_tokens=
|
238 |
call_agent=False, call_agent_level=call_agent_level)
|
239 |
-
call_result
|
240 |
-
|
|
|
|
|
241 |
else:
|
242 |
-
call_result = "Error:
|
243 |
else:
|
244 |
-
call_result = self.tooluniverse.run_one_function(
|
245 |
-
function_call_json[i])
|
246 |
-
|
247 |
call_id = self.tooluniverse.call_id_gen()
|
248 |
function_call_json[i]["call_id"] = call_id
|
249 |
-
|
250 |
call_results.append({
|
251 |
"role": "tool",
|
252 |
-
"content": json.dumps({"content": call_result, "call_id": call_id})
|
253 |
})
|
254 |
else:
|
255 |
call_results.append({
|
256 |
"role": "tool",
|
257 |
-
"content": json.dumps({"content": "
|
258 |
})
|
259 |
|
260 |
revised_messages = [{
|
@@ -262,8 +231,6 @@ class TxAgent:
|
|
262 |
"content": message.strip(),
|
263 |
"tool_calls": json.dumps(function_call_json)
|
264 |
}] + call_results
|
265 |
-
|
266 |
-
# Yield the final result.
|
267 |
return revised_messages, existing_tools_prompt, special_tool_call
|
268 |
|
269 |
def run_function_call_stream(self, fcall_str,
|
@@ -274,26 +241,24 @@ class TxAgent:
|
|
274 |
call_agent_level=None,
|
275 |
temperature=None,
|
276 |
return_gradio_history=True):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
277 |
|
278 |
-
function_call_json, message = self.tooluniverse.extract_function_call_json(
|
279 |
-
fcall_str, return_message=return_message, verbose=False)
|
280 |
call_results = []
|
281 |
special_tool_call = ''
|
282 |
if return_gradio_history:
|
283 |
gradio_history = []
|
284 |
-
if function_call_json
|
285 |
if isinstance(function_call_json, list):
|
286 |
for i in range(len(function_call_json)):
|
287 |
if function_call_json[i]["name"] == 'Finish':
|
288 |
special_tool_call = 'Finish'
|
289 |
break
|
290 |
-
elif function_call_json[i]["name"] == 'Tool_RAG':
|
291 |
-
new_tools_prompt, call_result = self.tool_RAG(
|
292 |
-
message=message,
|
293 |
-
existing_tools_prompt=existing_tools_prompt,
|
294 |
-
rag_num=self.step_rag_num,
|
295 |
-
return_call_result=True)
|
296 |
-
existing_tools_prompt += new_tools_prompt
|
297 |
elif function_call_json[i]["name"] == 'DirectResponse':
|
298 |
call_result = function_call_json[i]['arguments']['respose']
|
299 |
special_tool_call = 'DirectResponse'
|
@@ -308,42 +273,33 @@ class TxAgent:
|
|
308 |
"\nYou must follow the following plan to answer the question: " +
|
309 |
str(solution_plan)
|
310 |
)
|
311 |
-
sub_agent_task = "Sub TxAgent plan: " +
|
312 |
-
str(solution_plan)
|
313 |
-
# When streaming, yield responses as they arrive.
|
314 |
call_result = yield from self.run_gradio_chat(
|
315 |
full_message, history=[], temperature=temperature,
|
316 |
-
max_new_tokens=
|
317 |
call_agent=False, call_agent_level=call_agent_level,
|
318 |
-
conversation=None,
|
319 |
-
|
320 |
-
|
321 |
-
|
322 |
-
|
323 |
else:
|
324 |
-
call_result = "Error:
|
325 |
else:
|
326 |
-
call_result = self.tooluniverse.run_one_function(
|
327 |
-
function_call_json[i])
|
328 |
-
|
329 |
call_id = self.tooluniverse.call_id_gen()
|
330 |
function_call_json[i]["call_id"] = call_id
|
331 |
call_results.append({
|
332 |
"role": "tool",
|
333 |
-
"content": json.dumps({"content": call_result, "call_id": call_id})
|
334 |
})
|
335 |
if return_gradio_history and function_call_json[i]["name"] != 'Finish':
|
336 |
-
|
337 |
-
|
338 |
-
"title": "🧰 "+function_call_json[i]['name'], "log": str(function_call_json[i]['arguments'])}))
|
339 |
-
|
340 |
-
else:
|
341 |
-
gradio_history.append(ChatMessage(role="assistant", content=str(call_result), metadata={
|
342 |
-
"title": "⚒️ "+function_call_json[i]['name'], "log": str(function_call_json[i]['arguments'])}))
|
343 |
else:
|
344 |
call_results.append({
|
345 |
"role": "tool",
|
346 |
-
"content": json.dumps({"content": "
|
347 |
})
|
348 |
|
349 |
revised_messages = [{
|
@@ -351,152 +307,119 @@ class TxAgent:
|
|
351 |
"content": message.strip(),
|
352 |
"tool_calls": json.dumps(function_call_json)
|
353 |
}] + call_results
|
354 |
-
|
355 |
-
# Yield the final result.
|
356 |
if return_gradio_history:
|
357 |
return revised_messages, existing_tools_prompt, special_tool_call, gradio_history
|
358 |
-
|
359 |
-
return revised_messages, existing_tools_prompt, special_tool_call
|
360 |
|
361 |
def get_answer_based_on_unfinished_reasoning(self, conversation, temperature, max_new_tokens, max_token, outputs=None):
|
362 |
-
if conversation[-1]['role'] == '
|
363 |
conversation.append(
|
364 |
-
{'role': 'tool', 'content': 'Errors
|
365 |
finish_tools_prompt = self.add_finish_tools([])
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
-
|
370 |
-
|
371 |
-
|
372 |
-
|
373 |
-
|
|
|
374 |
return last_outputs_str
|
375 |
|
376 |
def run_multistep_agent(self, message: str,
|
377 |
temperature: float,
|
378 |
max_new_tokens: int,
|
379 |
max_token: int,
|
380 |
-
max_round: int =
|
381 |
call_agent=False,
|
382 |
-
call_agent_level=0)
|
383 |
-
""
|
384 |
-
Generate a streaming response using the llama3-8b model.
|
385 |
-
Args:
|
386 |
-
message (str): The input message.
|
387 |
-
temperature (float): The temperature for generating the response.
|
388 |
-
max_new_tokens (int): The maximum number of new tokens to generate.
|
389 |
-
Returns:
|
390 |
-
str: The generated response.
|
391 |
-
"""
|
392 |
-
print("\033[1;32;40mstart\033[0m")
|
393 |
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
394 |
call_agent, call_agent_level, message)
|
395 |
conversation = self.initialize_conversation(message)
|
396 |
-
|
397 |
outputs = []
|
398 |
last_outputs = []
|
399 |
next_round = True
|
400 |
-
function_call_messages = []
|
401 |
current_round = 0
|
402 |
token_overflow = False
|
403 |
enable_summary = False
|
404 |
last_status = {}
|
405 |
|
406 |
-
|
407 |
-
|
408 |
-
|
409 |
-
|
410 |
-
|
411 |
-
|
412 |
-
|
413 |
-
|
414 |
-
|
415 |
-
|
416 |
-
|
417 |
-
|
418 |
-
temperature=temperature)
|
419 |
-
|
420 |
-
if special_tool_call == 'Finish':
|
421 |
-
next_round = False
|
422 |
-
conversation.extend(function_call_messages)
|
423 |
-
if isinstance(function_call_messages[0]['content'], types.GeneratorType):
|
424 |
-
function_call_messages[0]['content'] = next(
|
425 |
-
function_call_messages[0]['content'])
|
426 |
-
return function_call_messages[0]['content'].split('[FinalAnswer]')[-1]
|
427 |
-
|
428 |
-
if (self.enable_summary or token_overflow) and not call_agent:
|
429 |
-
if token_overflow:
|
430 |
-
print("token_overflow, using summary")
|
431 |
-
enable_summary = True
|
432 |
-
last_status = self.function_result_summary(
|
433 |
-
conversation, status=last_status, enable_summary=enable_summary)
|
434 |
-
|
435 |
-
if function_call_messages is not None:
|
436 |
-
conversation.extend(function_call_messages)
|
437 |
-
outputs.append(tool_result_format(
|
438 |
-
function_call_messages))
|
439 |
-
else:
|
440 |
-
next_round = False
|
441 |
-
conversation.extend(
|
442 |
-
[{"role": "assistant", "content": ''.join(last_outputs)}])
|
443 |
-
return ''.join(last_outputs).replace("</s>", "")
|
444 |
-
if self.enable_checker:
|
445 |
-
good_status, wrong_info = checker.check_conversation()
|
446 |
-
if not good_status:
|
447 |
-
next_round = False
|
448 |
-
print(
|
449 |
-
"Internal error in reasoning: " + wrong_info)
|
450 |
-
break
|
451 |
-
last_outputs = []
|
452 |
-
outputs.append("### TxAgent:\n")
|
453 |
-
last_outputs_str, token_overflow = self.llm_infer(messages=conversation,
|
454 |
-
temperature=temperature,
|
455 |
-
tools=picked_tools_prompt,
|
456 |
-
skip_special_tokens=False,
|
457 |
-
max_new_tokens=max_new_tokens, max_token=max_token,
|
458 |
-
check_token_status=True)
|
459 |
-
if last_outputs_str is None:
|
460 |
next_round = False
|
461 |
-
|
462 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
463 |
else:
|
464 |
-
|
465 |
-
|
466 |
-
|
467 |
-
|
468 |
-
|
469 |
-
|
470 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
471 |
|
472 |
-
|
473 |
-
|
474 |
-
|
475 |
-
|
476 |
-
|
477 |
-
|
478 |
|
479 |
def build_logits_processor(self, messages, llm):
|
480 |
-
#
|
481 |
-
|
482 |
-
|
483 |
-
|
484 |
-
|
485 |
-
|
486 |
-
|
487 |
-
|
488 |
-
|
489 |
-
|
490 |
-
msg, add_special_tokens=False) for msg in assistant_messages]
|
491 |
-
return [NoRepeatSentenceProcessor(forbidden_ids, 5)]
|
492 |
-
else:
|
493 |
-
return None
|
494 |
|
495 |
def llm_infer(self, messages, temperature=0.1, tools=None,
|
496 |
-
output_begin_string=None, max_new_tokens=
|
497 |
-
max_token=
|
498 |
-
model=None, tokenizer=None, terminators=None,
|
499 |
-
|
500 |
if model is None:
|
501 |
model = self.model
|
502 |
|
@@ -504,7 +427,6 @@ class TxAgent:
|
|
504 |
sampling_params = SamplingParams(
|
505 |
temperature=temperature,
|
506 |
max_tokens=max_new_tokens,
|
507 |
-
logits_processors=logits_processor,
|
508 |
seed=seed if seed is not None else self.seed,
|
509 |
)
|
510 |
|
@@ -515,244 +437,174 @@ class TxAgent:
|
|
515 |
|
516 |
if check_token_status and max_token is not None:
|
517 |
token_overflow = False
|
518 |
-
num_input_tokens = len(self.tokenizer.encode(
|
519 |
-
|
520 |
-
|
521 |
-
|
522 |
-
|
523 |
-
|
524 |
-
|
525 |
-
|
526 |
-
logger.info(
|
527 |
-
"The number of tokens exceeds the maximum limit!!!!")
|
528 |
-
token_overflow = True
|
529 |
-
return None, token_overflow
|
530 |
-
output = model.generate(
|
531 |
-
prompt,
|
532 |
-
sampling_params=sampling_params,
|
533 |
-
)
|
534 |
output = output[0].outputs[0].text
|
535 |
-
|
|
|
|
|
536 |
if check_token_status and max_token is not None:
|
537 |
return output, token_overflow
|
538 |
-
|
539 |
return output
|
540 |
|
541 |
def run_self_agent(self, message: str,
|
542 |
temperature: float,
|
543 |
max_new_tokens: int,
|
544 |
-
max_token: int)
|
545 |
-
|
546 |
-
|
547 |
-
conversation = []
|
548 |
-
conversation = self.set_system_prompt(conversation, self.self_prompt)
|
549 |
conversation.append({"role": "user", "content": message})
|
550 |
-
return self.llm_infer(
|
551 |
-
|
552 |
-
|
553 |
-
|
|
|
|
|
554 |
|
555 |
def run_chat_agent(self, message: str,
|
556 |
temperature: float,
|
557 |
max_new_tokens: int,
|
558 |
-
max_token: int)
|
559 |
-
|
560 |
-
|
561 |
-
conversation = []
|
562 |
-
conversation = self.set_system_prompt(conversation, self.chat_prompt)
|
563 |
conversation.append({"role": "user", "content": message})
|
564 |
-
return self.llm_infer(
|
565 |
-
|
566 |
-
|
567 |
-
|
|
|
|
|
568 |
|
569 |
def run_format_agent(self, message: str,
|
570 |
answer: str,
|
571 |
temperature: float,
|
572 |
max_new_tokens: int,
|
573 |
-
max_token: int)
|
574 |
-
|
575 |
-
print("\033[1;32;40mstart format agent\033[0m")
|
576 |
if '[FinalAnswer]' in answer:
|
577 |
possible_final_answer = answer.split("[FinalAnswer]")[-1]
|
578 |
elif "\n\n" in answer:
|
579 |
possible_final_answer = answer.split("\n\n")[-1]
|
580 |
else:
|
581 |
possible_final_answer = answer.strip()
|
582 |
-
if len(possible_final_answer) == 1:
|
583 |
-
|
584 |
-
|
585 |
-
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
-
if choice in ['A', 'B', 'C', 'D', 'E']:
|
590 |
-
print("choice", choice)
|
591 |
-
return choice
|
592 |
-
|
593 |
-
conversation = []
|
594 |
-
format_prompt = f"You are helpful assistant to transform the answer of agent to the final answer of 'A', 'B', 'C', 'D'."
|
595 |
-
conversation = self.set_system_prompt(conversation, format_prompt)
|
596 |
conversation.append({"role": "user", "content": message +
|
597 |
-
"\
|
598 |
-
return self.llm_infer(
|
599 |
-
|
600 |
-
|
601 |
-
|
|
|
|
|
602 |
|
603 |
def run_summary_agent(self, thought_calls: str,
|
604 |
function_response: str,
|
605 |
temperature: float,
|
606 |
max_new_tokens: int,
|
607 |
-
max_token: int)
|
608 |
-
|
609 |
-
|
610 |
{thought_calls}
|
611 |
-
|
612 |
Function calls' responses:
|
613 |
\"\"\"
|
614 |
{function_response}
|
615 |
\"\"\"
|
616 |
-
|
617 |
-
|
618 |
-
|
619 |
-
|
620 |
-
|
621 |
-
|
622 |
-
|
623 |
-
|
624 |
-
|
625 |
-
{"role": "user", "content": generate_tool_result_summary_training_prompt})
|
626 |
-
output = self.llm_infer(messages=conversation,
|
627 |
-
temperature=temperature,
|
628 |
-
tools=None,
|
629 |
-
max_new_tokens=max_new_tokens, max_token=max_token)
|
630 |
-
|
631 |
if '[' in output:
|
632 |
output = output.split('[')[0]
|
633 |
return output
|
634 |
|
635 |
def function_result_summary(self, input_list, status, enable_summary):
|
636 |
-
"""
|
637 |
-
Processes the input list, extracting information from sequences of 'user', 'tool', 'assistant' roles.
|
638 |
-
Supports 'length' and 'step' modes, and skips the last 'k' groups.
|
639 |
-
|
640 |
-
Parameters:
|
641 |
-
input_list (list): A list of dictionaries containing role and other information.
|
642 |
-
summary_skip_last_k (int): Number of groups to skip from the end. Defaults to 0.
|
643 |
-
summary_context_length (int): The context length threshold for the 'length' mode.
|
644 |
-
last_processed_index (tuple or int): The last processed index.
|
645 |
-
|
646 |
-
Returns:
|
647 |
-
list: A list of extracted information from valid sequences.
|
648 |
-
"""
|
649 |
if 'tool_call_step' not in status:
|
650 |
status['tool_call_step'] = 0
|
651 |
-
|
652 |
for idx in range(len(input_list)):
|
653 |
-
pos_id = len(input_list)-idx-1
|
654 |
-
if input_list[pos_id]['role'] == 'assistant':
|
655 |
-
if 'tool_calls' in input_list[pos_id]:
|
656 |
-
if 'Tool_RAG' in str(input_list[pos_id]['tool_calls']):
|
657 |
-
status['tool_call_step'] += 1
|
658 |
break
|
659 |
|
660 |
-
|
661 |
-
status['step'] += 1
|
662 |
-
else:
|
663 |
-
status['step'] = 0
|
664 |
-
|
665 |
if not enable_summary:
|
666 |
return status
|
667 |
|
668 |
-
|
669 |
-
|
670 |
-
|
671 |
-
|
672 |
-
status['summarized_step'] = 0
|
673 |
-
|
674 |
-
if 'previous_length' not in status:
|
675 |
-
status['previous_length'] = 0
|
676 |
-
|
677 |
-
if 'history' not in status:
|
678 |
-
status['history'] = []
|
679 |
|
680 |
function_response = ''
|
681 |
-
idx =
|
682 |
-
|
683 |
-
|
684 |
-
status['history'].append(self.summary_mode == 'step' and status['summarized_step']
|
685 |
-
< status['step']-status['tool_call_step']-self.summary_skip_last_k)
|
686 |
|
687 |
-
idx = current_summarized_index
|
688 |
while idx < len(input_list):
|
689 |
-
if (self.summary_mode == 'step' and status['summarized_step'] < status['step']-status['tool_call_step']-self.summary_skip_last_k) or
|
690 |
-
|
691 |
if input_list[idx]['role'] == 'assistant':
|
692 |
-
if
|
693 |
-
|
694 |
-
|
695 |
-
|
696 |
-
|
697 |
-
|
698 |
-
|
699 |
-
|
700 |
-
|
701 |
-
|
702 |
-
|
703 |
-
|
704 |
-
|
705 |
-
|
706 |
-
input_list.insert(
|
707 |
-
last_call_idx+1, {'role': 'tool', 'content': result_summary})
|
708 |
-
status['summarized_index'] = last_call_idx + 2
|
709 |
-
idx += 1
|
710 |
-
|
711 |
-
last_call_idx = idx
|
712 |
-
this_thought_calls = input_list[idx]['content'] + \
|
713 |
-
input_list[idx]['tool_calls']
|
714 |
-
function_response = ''
|
715 |
-
|
716 |
elif input_list[idx]['role'] == 'tool' and this_thought_calls is not None:
|
717 |
function_response += input_list[idx]['content']
|
718 |
del input_list[idx]
|
719 |
idx -= 1
|
720 |
-
|
721 |
else:
|
722 |
break
|
723 |
idx += 1
|
724 |
|
725 |
-
if
|
726 |
status['summarized_step'] += 1
|
727 |
result_summary = self.run_summary_agent(
|
728 |
thought_calls=this_thought_calls,
|
729 |
function_response=function_response,
|
730 |
temperature=0.1,
|
731 |
-
max_new_tokens=
|
732 |
-
max_token=
|
733 |
-
)
|
734 |
-
|
735 |
tool_calls = json.loads(input_list[last_call_idx]['tool_calls'])
|
736 |
for tool_call in tool_calls:
|
737 |
del tool_call['call_id']
|
738 |
input_list[last_call_idx]['tool_calls'] = json.dumps(tool_calls)
|
739 |
-
input_list.insert(
|
740 |
-
last_call_idx+1, {'role': 'tool', 'content': result_summary})
|
741 |
status['summarized_index'] = last_call_idx + 2
|
742 |
|
743 |
return status
|
744 |
|
745 |
-
# Following are Gradio related functions
|
746 |
-
|
747 |
-
# General update method that accepts any new arguments through kwargs
|
748 |
def update_parameters(self, **kwargs):
|
|
|
749 |
for key, value in kwargs.items():
|
750 |
if hasattr(self, key):
|
751 |
setattr(self, key, value)
|
752 |
-
|
753 |
-
|
754 |
-
updated_attributes = {key: value for key,
|
755 |
-
value in kwargs.items() if hasattr(self, key)}
|
756 |
return updated_attributes
|
757 |
|
758 |
def run_gradio_chat(self, message: str,
|
@@ -762,54 +614,33 @@ Generate **one summarized sentence** about "function calls' responses" with nece
|
|
762 |
max_token: int,
|
763 |
call_agent: bool,
|
764 |
conversation: gr.State,
|
765 |
-
max_round: int =
|
766 |
seed: int = None,
|
767 |
call_agent_level: int = 0,
|
768 |
-
sub_agent_task: str = None
|
769 |
-
|
770 |
-
|
771 |
-
|
772 |
-
message
|
773 |
-
|
774 |
-
temperature (float): The temperature for generating the response.
|
775 |
-
max_new_tokens (int): The maximum number of new tokens to generate.
|
776 |
-
Returns:
|
777 |
-
str: The generated response.
|
778 |
-
"""
|
779 |
-
print("\033[1;32;40mstart\033[0m")
|
780 |
-
print("len(message)", len(message))
|
781 |
-
if len(message) <= 10:
|
782 |
-
yield "Hi, I am TxAgent, an assistant for answering biomedical questions. Please provide a valid message with a string longer than 10 characters."
|
783 |
-
return "Please provide a valid message."
|
784 |
-
outputs = []
|
785 |
-
outputs_str = ''
|
786 |
-
last_outputs = []
|
787 |
|
788 |
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
789 |
-
call_agent,
|
790 |
-
call_agent_level,
|
791 |
-
message)
|
792 |
-
|
793 |
conversation = self.initialize_conversation(
|
794 |
-
message,
|
795 |
-
conversation=conversation,
|
796 |
-
history=history)
|
797 |
history = []
|
|
|
798 |
|
799 |
next_round = True
|
800 |
-
function_call_messages = []
|
801 |
current_round = 0
|
802 |
enable_summary = False
|
803 |
-
last_status = {}
|
804 |
token_overflow = False
|
805 |
-
if self.enable_checker:
|
806 |
-
checker = ReasoningTraceChecker(
|
807 |
-
message, conversation, init_index=len(conversation))
|
808 |
|
809 |
try:
|
810 |
while next_round and current_round < max_round:
|
811 |
current_round += 1
|
812 |
-
if
|
813 |
function_call_messages, picked_tools_prompt, special_tool_call, current_gradio_history = yield from self.run_function_call_stream(
|
814 |
last_outputs, return_message=True,
|
815 |
existing_tools_prompt=picked_tools_prompt,
|
@@ -818,40 +649,33 @@ Generate **one summarized sentence** about "function calls' responses" with nece
|
|
818 |
call_agent_level=call_agent_level,
|
819 |
temperature=temperature)
|
820 |
history.extend(current_gradio_history)
|
|
|
821 |
if special_tool_call == 'Finish':
|
822 |
yield history
|
823 |
next_round = False
|
824 |
conversation.extend(function_call_messages)
|
825 |
return function_call_messages[0]['content']
|
826 |
-
|
827 |
-
|
828 |
-
|
|
|
829 |
yield history
|
830 |
next_round = False
|
831 |
-
return
|
|
|
832 |
if (self.enable_summary or token_overflow) and not call_agent:
|
833 |
-
if token_overflow:
|
834 |
-
print("token_overflow, using summary")
|
835 |
enable_summary = True
|
836 |
last_status = self.function_result_summary(
|
837 |
-
conversation, status=last_status,
|
838 |
-
|
839 |
-
if function_call_messages
|
840 |
conversation.extend(function_call_messages)
|
841 |
-
formated_md_function_call_messages = tool_result_format(
|
842 |
-
function_call_messages)
|
843 |
yield history
|
844 |
else:
|
845 |
next_round = False
|
846 |
-
conversation.
|
847 |
-
[{"role": "assistant", "content": ''.join(last_outputs)}])
|
848 |
return ''.join(last_outputs).replace("</s>", "")
|
849 |
-
|
850 |
-
good_status, wrong_info = checker.check_conversation()
|
851 |
-
if not good_status:
|
852 |
-
next_round = False
|
853 |
-
print("Internal error in reasoning: " + wrong_info)
|
854 |
-
break
|
855 |
last_outputs = []
|
856 |
last_outputs_str, token_overflow = self.llm_infer(
|
857 |
messages=conversation,
|
@@ -862,26 +686,34 @@ Generate **one summarized sentence** about "function calls' responses" with nece
|
|
862 |
max_token=max_token,
|
863 |
seed=seed,
|
864 |
check_token_status=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
865 |
last_thought = last_outputs_str.split("[TOOL_CALLS]")[0]
|
866 |
-
for
|
867 |
-
if
|
868 |
-
|
|
|
869 |
if '[FinalAnswer]' in last_thought:
|
870 |
-
|
871 |
-
|
872 |
-
history.append(
|
873 |
-
ChatMessage(role="assistant",
|
874 |
-
content=final_thought.strip())
|
875 |
-
)
|
876 |
yield history
|
877 |
-
history.append(
|
878 |
-
ChatMessage(
|
879 |
-
role="assistant", content="**Answer**:\n"+final_answer.strip())
|
880 |
-
)
|
881 |
yield history
|
882 |
else:
|
883 |
-
history.append(ChatMessage(
|
884 |
-
role="assistant", content=last_thought))
|
885 |
yield history
|
886 |
|
887 |
last_outputs.append(last_outputs_str)
|
@@ -890,48 +722,27 @@ Generate **one summarized sentence** about "function calls' responses" with nece
|
|
890 |
if self.force_finish:
|
891 |
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
892 |
conversation, temperature, max_new_tokens, max_token)
|
893 |
-
|
894 |
-
|
895 |
-
|
896 |
-
|
897 |
-
|
898 |
-
|
899 |
-
history.append(
|
900 |
-
ChatMessage(role="assistant",
|
901 |
-
content=final_thought.strip())
|
902 |
-
)
|
903 |
-
yield history
|
904 |
-
history.append(
|
905 |
-
ChatMessage(
|
906 |
-
role="assistant", content="**Answer**:\n"+final_answer.strip())
|
907 |
-
)
|
908 |
-
yield history
|
909 |
else:
|
910 |
-
yield "
|
911 |
|
912 |
except Exception as e:
|
913 |
-
|
|
|
|
|
|
|
914 |
if self.force_finish:
|
915 |
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
916 |
-
conversation,
|
917 |
-
|
918 |
-
|
919 |
-
|
920 |
-
|
921 |
-
|
922 |
-
|
923 |
-
|
924 |
-
final_thought, final_answer = last_thought.split(
|
925 |
-
'[FinalAnswer]')
|
926 |
-
history.append(
|
927 |
-
ChatMessage(role="assistant",
|
928 |
-
content=final_thought.strip())
|
929 |
-
)
|
930 |
-
yield history
|
931 |
-
history.append(
|
932 |
-
ChatMessage(
|
933 |
-
role="assistant", content="**Answer**:\n"+final_answer.strip())
|
934 |
-
)
|
935 |
-
yield history
|
936 |
-
else:
|
937 |
-
return None
|
|
|
11 |
from tooluniverse import ToolUniverse
|
12 |
from gradio import ChatMessage
|
13 |
from .toolrag import ToolRAGModel
|
14 |
+
import torch
|
15 |
+
import logging
|
16 |
|
17 |
+
logger = logging.getLogger(__name__)
|
18 |
+
logging.basicConfig(level=logging.INFO)
|
19 |
|
20 |
+
from .utils import NoRepeatSentenceProcessor, ReasoningTraceChecker, tool_result_format
|
21 |
|
22 |
class TxAgent:
|
23 |
def __init__(self, model_name,
|
24 |
rag_model_name,
|
25 |
+
tool_files_dict=None,
|
26 |
enable_finish=True,
|
27 |
+
enable_rag=False,
|
28 |
enable_summary=False,
|
29 |
init_rag_num=0,
|
30 |
+
step_rag_num=0,
|
31 |
summary_mode='step',
|
32 |
summary_skip_last_k=0,
|
33 |
summary_context_length=None,
|
|
|
36 |
seed=None,
|
37 |
enable_checker=False,
|
38 |
enable_chat=False,
|
39 |
+
additional_default_tools=None):
|
|
|
40 |
self.model_name = model_name
|
41 |
self.tokenizer = None
|
42 |
self.terminators = None
|
|
|
45 |
self.model = None
|
46 |
self.rag_model = ToolRAGModel(rag_model_name)
|
47 |
self.tooluniverse = None
|
48 |
+
self.prompt_multi_step = "You are a helpful assistant that solves problems through step-by-step reasoning."
|
|
|
49 |
self.self_prompt = "Strictly follow the instruction."
|
50 |
+
self.chat_prompt = "You are a helpful assistant for user chat."
|
51 |
self.enable_finish = enable_finish
|
52 |
self.enable_rag = enable_rag
|
53 |
self.enable_summary = enable_summary
|
|
|
61 |
self.seed = seed
|
62 |
self.enable_checker = enable_checker
|
63 |
self.additional_default_tools = additional_default_tools
|
64 |
+
logger.info("TxAgent initialized with model: %s, RAG: %s", model_name, rag_model_name)
|
65 |
|
66 |
def init_model(self):
|
67 |
self.load_models()
|
68 |
self.load_tooluniverse()
|
|
|
|
|
|
|
|
|
|
|
69 |
|
70 |
def load_models(self, model_name=None):
|
71 |
if model_name is not None:
|
|
|
73 |
return f"The model {model_name} is already loaded."
|
74 |
self.model_name = model_name
|
75 |
|
76 |
+
self.model = LLM(model=self.model_name, dtype="float16", max_model_len=2048, gpu_memory_utilization=0.8)
|
77 |
self.chat_template = Template(self.model.get_tokenizer().chat_template)
|
78 |
self.tokenizer = self.model.get_tokenizer()
|
79 |
+
logger.info("Model %s loaded successfully", self.model_name)
|
80 |
return f"Model {model_name} loaded successfully."
|
81 |
|
82 |
def load_tooluniverse(self):
|
|
|
85 |
special_tools = self.tooluniverse.prepare_tool_prompts(
|
86 |
self.tooluniverse.tool_category_dicts["special_tools"])
|
87 |
self.special_tools_name = [tool['name'] for tool in special_tools]
|
88 |
+
logger.debug("ToolUniverse loaded with %d special tools", len(self.special_tools_name))
|
89 |
|
90 |
def load_tool_desc_embedding(self):
|
91 |
+
cache_path = os.path.join(os.path.dirname(self.tool_files_dict["new_tool"]), "tool_embeddings.pkl")
|
92 |
+
if os.path.exists(cache_path):
|
93 |
+
self.rag_model.load_cached_embeddings(cache_path)
|
94 |
+
else:
|
95 |
+
self.rag_model.load_tool_desc_embedding(self.tooluniverse)
|
96 |
+
self.rag_model.save_embeddings(cache_path)
|
97 |
+
logger.debug("Tool description embeddings loaded")
|
98 |
|
99 |
def rag_infer(self, query, top_k=5):
|
100 |
return self.rag_model.rag_infer(query, top_k)
|
|
|
107 |
call_agent_level += 1
|
108 |
if call_agent_level >= 2:
|
109 |
call_agent = False
|
|
|
|
|
|
|
|
|
110 |
return picked_tools_prompt, call_agent_level
|
111 |
|
112 |
def initialize_conversation(self, message, conversation=None, history=None):
|
|
|
115 |
|
116 |
conversation = self.set_system_prompt(
|
117 |
conversation, self.prompt_multi_step)
|
118 |
+
if history:
|
119 |
+
for i in range(len(history)):
|
120 |
+
if history[i]['role'] == 'user':
|
121 |
+
conversation.append({"role": "user", "content": history[i]['content']})
|
122 |
+
elif history[i]['role'] == 'assistant':
|
123 |
+
conversation.append({"role": "assistant", "content": history[i]['content']})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
124 |
conversation.append({"role": "user", "content": message})
|
125 |
+
logger.debug("Conversation initialized with %d messages", len(conversation))
|
126 |
return conversation
|
127 |
|
128 |
def tool_RAG(self, message=None,
|
129 |
picked_tool_names=None,
|
130 |
existing_tools_prompt=[],
|
131 |
+
rag_num=0,
|
132 |
return_call_result=False):
|
133 |
+
if not self.enable_rag:
|
134 |
+
return []
|
135 |
+
extra_factor = 10
|
136 |
if picked_tool_names is None:
|
137 |
assert picked_tool_names is not None or message is not None
|
138 |
picked_tool_names = self.rag_infer(
|
139 |
+
message, top_k=rag_num * extra_factor)
|
140 |
|
141 |
+
picked_tool_names_no_special = [tool for tool in picked_tool_names if tool not in self.special_tools_name]
|
|
|
|
|
|
|
|
|
142 |
picked_tool_names = picked_tool_names_no_special[:rag_num]
|
143 |
|
144 |
picked_tools = self.tooluniverse.get_tool_by_name(picked_tool_names)
|
145 |
+
picked_tools_prompt = self.tooluniverse.prepare_tool_prompts(picked_tools)
|
146 |
+
logger.debug("Retrieved %d tools via RAG", len(picked_tools_prompt))
|
147 |
if return_call_result:
|
148 |
return picked_tools_prompt, picked_tool_names
|
149 |
return picked_tools_prompt
|
150 |
|
151 |
def add_special_tools(self, tools, call_agent=False):
|
152 |
if self.enable_finish:
|
153 |
+
tools.append(self.tooluniverse.get_one_tool_by_one_name('Finish', return_prompt=True))
|
154 |
+
logger.debug("Finish tool added")
|
|
|
155 |
if call_agent:
|
156 |
+
tools.append(self.tooluniverse.get_one_tool_by_one_name('CallAgent', return_prompt=True))
|
157 |
+
logger.debug("CallAgent tool added")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
158 |
return tools
|
159 |
|
160 |
def add_finish_tools(self, tools):
|
161 |
+
tools.append(self.tooluniverse.get_one_tool_by_one_name('Finish', return_prompt=True))
|
162 |
+
logger.debug("Finish tool added")
|
|
|
163 |
return tools
|
164 |
|
165 |
def set_system_prompt(self, conversation, sys_prompt):
|
166 |
+
if not conversation:
|
167 |
+
conversation.append({"role": "system", "content": sys_prompt})
|
|
|
168 |
else:
|
169 |
conversation[0] = {"role": "system", "content": sys_prompt}
|
170 |
return conversation
|
|
|
176 |
call_agent=False,
|
177 |
call_agent_level=None,
|
178 |
temperature=None):
|
179 |
+
try:
|
180 |
+
function_call_json, message = self.tooluniverse.extract_function_call_json(
|
181 |
+
fcall_str, return_message=return_message, verbose=False)
|
182 |
+
except Exception as e:
|
183 |
+
logger.error("Tool call parsing failed: %s", e)
|
184 |
+
function_call_json = []
|
185 |
+
message = fcall_str
|
186 |
|
|
|
|
|
187 |
call_results = []
|
188 |
special_tool_call = ''
|
189 |
+
if function_call_json:
|
190 |
if isinstance(function_call_json, list):
|
191 |
for i in range(len(function_call_json)):
|
192 |
+
logger.info("Tool Call: %s", function_call_json[i])
|
193 |
if function_call_json[i]["name"] == 'Finish':
|
194 |
special_tool_call = 'Finish'
|
195 |
break
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
196 |
elif function_call_json[i]["name"] == 'CallAgent':
|
197 |
if call_agent_level < 2 and call_agent:
|
198 |
solution_plan = function_call_json[i]['arguments']['solution']
|
|
|
203 |
)
|
204 |
call_result = self.run_multistep_agent(
|
205 |
full_message, temperature=temperature,
|
206 |
+
max_new_tokens=512, max_token=2048,
|
207 |
call_agent=False, call_agent_level=call_agent_level)
|
208 |
+
if call_result is None:
|
209 |
+
call_result = "⚠️ No content returned from sub-agent."
|
210 |
+
else:
|
211 |
+
call_result = call_result.split('[FinalAnswer]')[-1].strip()
|
212 |
else:
|
213 |
+
call_result = "Error: CallAgent disabled."
|
214 |
else:
|
215 |
+
call_result = self.tooluniverse.run_one_function(function_call_json[i])
|
|
|
|
|
216 |
call_id = self.tooluniverse.call_id_gen()
|
217 |
function_call_json[i]["call_id"] = call_id
|
218 |
+
logger.info("Tool Call Result: %s", call_result)
|
219 |
call_results.append({
|
220 |
"role": "tool",
|
221 |
+
"content": json.dumps({"tool_name": function_call_json[i]["name"], "content": call_result, "call_id": call_id})
|
222 |
})
|
223 |
else:
|
224 |
call_results.append({
|
225 |
"role": "tool",
|
226 |
+
"content": json.dumps({"content": "Invalid or no function call detected."})
|
227 |
})
|
228 |
|
229 |
revised_messages = [{
|
|
|
231 |
"content": message.strip(),
|
232 |
"tool_calls": json.dumps(function_call_json)
|
233 |
}] + call_results
|
|
|
|
|
234 |
return revised_messages, existing_tools_prompt, special_tool_call
|
235 |
|
236 |
def run_function_call_stream(self, fcall_str,
|
|
|
241 |
call_agent_level=None,
|
242 |
temperature=None,
|
243 |
return_gradio_history=True):
|
244 |
+
try:
|
245 |
+
function_call_json, message = self.tooluniverse.extract_function_call_json(
|
246 |
+
fcall_str, return_message=return_message, verbose=False)
|
247 |
+
except Exception as e:
|
248 |
+
logger.error("Tool call parsing failed: %s", e)
|
249 |
+
function_call_json = []
|
250 |
+
message = fcall_str
|
251 |
|
|
|
|
|
252 |
call_results = []
|
253 |
special_tool_call = ''
|
254 |
if return_gradio_history:
|
255 |
gradio_history = []
|
256 |
+
if function_call_json:
|
257 |
if isinstance(function_call_json, list):
|
258 |
for i in range(len(function_call_json)):
|
259 |
if function_call_json[i]["name"] == 'Finish':
|
260 |
special_tool_call = 'Finish'
|
261 |
break
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
262 |
elif function_call_json[i]["name"] == 'DirectResponse':
|
263 |
call_result = function_call_json[i]['arguments']['respose']
|
264 |
special_tool_call = 'DirectResponse'
|
|
|
273 |
"\nYou must follow the following plan to answer the question: " +
|
274 |
str(solution_plan)
|
275 |
)
|
276 |
+
sub_agent_task = "Sub TxAgent plan: " + str(solution_plan)
|
|
|
|
|
277 |
call_result = yield from self.run_gradio_chat(
|
278 |
full_message, history=[], temperature=temperature,
|
279 |
+
max_new_tokens=512, max_token=2048,
|
280 |
call_agent=False, call_agent_level=call_agent_level,
|
281 |
+
conversation=None, sub_agent_task=sub_agent_task)
|
282 |
+
if call_result is not None and isinstance(call_result, str):
|
283 |
+
call_result = call_result.split('[FinalAnswer]')[-1]
|
284 |
+
else:
|
285 |
+
call_result = "⚠️ No content returned from sub-agent."
|
286 |
else:
|
287 |
+
call_result = "Error: CallAgent disabled."
|
288 |
else:
|
289 |
+
call_result = self.tooluniverse.run_one_function(function_call_json[i])
|
|
|
|
|
290 |
call_id = self.tooluniverse.call_id_gen()
|
291 |
function_call_json[i]["call_id"] = call_id
|
292 |
call_results.append({
|
293 |
"role": "tool",
|
294 |
+
"content": json.dumps({"tool_name": function_call_json[i]["name"], "content": call_result, "call_id": call_id})
|
295 |
})
|
296 |
if return_gradio_history and function_call_json[i]["name"] != 'Finish':
|
297 |
+
metadata = {"title": f"🧰 {function_call_json[i]['name']}", "log": str(function_call_json[i]['arguments'])}
|
298 |
+
gradio_history.append(ChatMessage(role="assistant", content=str(call_result), metadata=metadata))
|
|
|
|
|
|
|
|
|
|
|
299 |
else:
|
300 |
call_results.append({
|
301 |
"role": "tool",
|
302 |
+
"content": json.dumps({"content": "Invalid or no function call detected."})
|
303 |
})
|
304 |
|
305 |
revised_messages = [{
|
|
|
307 |
"content": message.strip(),
|
308 |
"tool_calls": json.dumps(function_call_json)
|
309 |
}] + call_results
|
|
|
|
|
310 |
if return_gradio_history:
|
311 |
return revised_messages, existing_tools_prompt, special_tool_call, gradio_history
|
312 |
+
return revised_messages, existing_tools_prompt, special_tool_call
|
|
|
313 |
|
314 |
def get_answer_based_on_unfinished_reasoning(self, conversation, temperature, max_new_tokens, max_token, outputs=None):
|
315 |
+
if conversation[-1]['role'] == 'assistant':
|
316 |
conversation.append(
|
317 |
+
{'role': 'tool', 'content': 'Errors occurred during function call; provide final answer with current information.'})
|
318 |
finish_tools_prompt = self.add_finish_tools([])
|
319 |
+
last_outputs_str = self.llm_infer(
|
320 |
+
messages=conversation,
|
321 |
+
temperature=temperature,
|
322 |
+
tools=finish_tools_prompt,
|
323 |
+
output_begin_string='[FinalAnswer]',
|
324 |
+
skip_special_tokens=True,
|
325 |
+
max_new_tokens=max_new_tokens,
|
326 |
+
max_token=max_token)
|
327 |
+
logger.info("Unfinished reasoning answer: %s", last_outputs_str[:100])
|
328 |
return last_outputs_str
|
329 |
|
330 |
def run_multistep_agent(self, message: str,
|
331 |
temperature: float,
|
332 |
max_new_tokens: int,
|
333 |
max_token: int,
|
334 |
+
max_round: int = 5,
|
335 |
call_agent=False,
|
336 |
+
call_agent_level=0):
|
337 |
+
logger.info("Starting multistep agent for message: %s", message[:100])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
338 |
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
339 |
call_agent, call_agent_level, message)
|
340 |
conversation = self.initialize_conversation(message)
|
|
|
341 |
outputs = []
|
342 |
last_outputs = []
|
343 |
next_round = True
|
|
|
344 |
current_round = 0
|
345 |
token_overflow = False
|
346 |
enable_summary = False
|
347 |
last_status = {}
|
348 |
|
349 |
+
while next_round and current_round < max_round:
|
350 |
+
current_round += 1
|
351 |
+
if len(outputs) > 0:
|
352 |
+
function_call_messages, picked_tools_prompt, special_tool_call = self.run_function_call(
|
353 |
+
last_outputs, return_message=True,
|
354 |
+
existing_tools_prompt=picked_tools_prompt,
|
355 |
+
message_for_call_agent=message,
|
356 |
+
call_agent=call_agent,
|
357 |
+
call_agent_level=call_agent_level,
|
358 |
+
temperature=temperature)
|
359 |
+
|
360 |
+
if special_tool_call == 'Finish':
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
361 |
next_round = False
|
362 |
+
conversation.extend(function_call_messages)
|
363 |
+
content = function_call_messages[0]['content']
|
364 |
+
if content is None:
|
365 |
+
return "❌ No content returned after Finish tool call."
|
366 |
+
return content.split('[FinalAnswer]')[-1]
|
367 |
+
|
368 |
+
if (self.enable_summary or token_overflow) and not call_agent:
|
369 |
+
enable_summary = True
|
370 |
+
last_status = self.function_result_summary(
|
371 |
+
conversation, status=last_status, enable_summary=enable_summary)
|
372 |
+
|
373 |
+
if function_call_messages:
|
374 |
+
conversation.extend(function_call_messages)
|
375 |
+
outputs.append(tool_result_format(function_call_messages))
|
376 |
else:
|
377 |
+
next_round = False
|
378 |
+
conversation.extend([{"role": "assistant", "content": ''.join(last_outputs)}])
|
379 |
+
return ''.join(last_outputs).replace("</s>", "")
|
380 |
+
|
381 |
+
last_outputs = []
|
382 |
+
outputs.append("### TxAgent:\n")
|
383 |
+
last_outputs_str, token_overflow = self.llm_infer(
|
384 |
+
messages=conversation,
|
385 |
+
temperature=temperature,
|
386 |
+
tools=picked_tools_prompt,
|
387 |
+
skip_special_tokens=False,
|
388 |
+
max_new_tokens=max_new_tokens,
|
389 |
+
max_token=max_token,
|
390 |
+
check_token_status=True)
|
391 |
+
if last_outputs_str is None:
|
392 |
+
logger.warning("Token limit exceeded")
|
393 |
+
if self.force_finish:
|
394 |
+
return self.get_answer_based_on_unfinished_reasoning(
|
395 |
+
conversation, temperature, max_new_tokens, max_token)
|
396 |
+
return "❌ Token limit exceeded."
|
397 |
+
last_outputs.append(last_outputs_str)
|
398 |
|
399 |
+
if max_round == current_round:
|
400 |
+
logger.warning("Max rounds exceeded")
|
401 |
+
if self.force_finish:
|
402 |
+
return self.get_answer_based_on_unfinished_reasoning(
|
403 |
+
conversation, temperature, max_new_tokens, max_token)
|
404 |
+
return None
|
405 |
|
406 |
def build_logits_processor(self, messages, llm):
|
407 |
+
# Disabled logits processor due to vLLM V1 limitation
|
408 |
+
logger.warning("Logits processor disabled due to vLLM V1 limitation")
|
409 |
+
return None
|
410 |
+
# Original code (commented out):
|
411 |
+
# tokenizer = llm.get_tokenizer()
|
412 |
+
# if self.avoid_repeat and len(messages) > 2:
|
413 |
+
# assistant_messages = [msg['content'] for msg in messages[-3:] if msg['role'] == 'assistant'][:2]
|
414 |
+
# forbidden_ids = [tokenizer.encode(msg, add_special_tokens=False) for msg in assistant_messages]
|
415 |
+
# return [NoRepeatSentenceProcessor(forbidden_ids, 5)]
|
416 |
+
# return None
|
|
|
|
|
|
|
|
|
417 |
|
418 |
def llm_infer(self, messages, temperature=0.1, tools=None,
|
419 |
+
output_begin_string=None, max_new_tokens=512,
|
420 |
+
max_token=2048, skip_special_tokens=True,
|
421 |
+
model=None, tokenizer=None, terminators=None,
|
422 |
+
seed=None, check_token_status=False):
|
423 |
if model is None:
|
424 |
model = self.model
|
425 |
|
|
|
427 |
sampling_params = SamplingParams(
|
428 |
temperature=temperature,
|
429 |
max_tokens=max_new_tokens,
|
|
|
430 |
seed=seed if seed is not None else self.seed,
|
431 |
)
|
432 |
|
|
|
437 |
|
438 |
if check_token_status and max_token is not None:
|
439 |
token_overflow = False
|
440 |
+
num_input_tokens = len(self.tokenizer.encode(prompt, return_tensors="pt")[0])
|
441 |
+
if num_input_tokens > max_token:
|
442 |
+
torch.cuda.empty_cache()
|
443 |
+
gc.collect()
|
444 |
+
logger.info("Token overflow: %d > %d", num_input_tokens, max_token)
|
445 |
+
return None, True
|
446 |
+
|
447 |
+
output = model.generate(prompt, sampling_params=sampling_params)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
448 |
output = output[0].outputs[0].text
|
449 |
+
logger.debug("Inference output: %s", output[:100])
|
450 |
+
torch.cuda.empty_cache()
|
451 |
+
gc.collect()
|
452 |
if check_token_status and max_token is not None:
|
453 |
return output, token_overflow
|
|
|
454 |
return output
|
455 |
|
456 |
def run_self_agent(self, message: str,
|
457 |
temperature: float,
|
458 |
max_new_tokens: int,
|
459 |
+
max_token: int):
|
460 |
+
logger.info("Starting self agent")
|
461 |
+
conversation = self.set_system_prompt([], self.self_prompt)
|
|
|
|
|
462 |
conversation.append({"role": "user", "content": message})
|
463 |
+
return self.llm_infer(
|
464 |
+
messages=conversation,
|
465 |
+
temperature=temperature,
|
466 |
+
tools=None,
|
467 |
+
max_new_tokens=max_new_tokens,
|
468 |
+
max_token=max_token)
|
469 |
|
470 |
def run_chat_agent(self, message: str,
|
471 |
temperature: float,
|
472 |
max_new_tokens: int,
|
473 |
+
max_token: int):
|
474 |
+
logger.info("Starting chat agent")
|
475 |
+
conversation = self.set_system_prompt([], self.chat_prompt)
|
|
|
|
|
476 |
conversation.append({"role": "user", "content": message})
|
477 |
+
return self.llm_infer(
|
478 |
+
messages=conversation,
|
479 |
+
temperature=temperature,
|
480 |
+
tools=None,
|
481 |
+
max_new_tokens=max_new_tokens,
|
482 |
+
max_token=max_token)
|
483 |
|
484 |
def run_format_agent(self, message: str,
|
485 |
answer: str,
|
486 |
temperature: float,
|
487 |
max_new_tokens: int,
|
488 |
+
max_token: int):
|
489 |
+
logger.info("Starting format agent")
|
|
|
490 |
if '[FinalAnswer]' in answer:
|
491 |
possible_final_answer = answer.split("[FinalAnswer]")[-1]
|
492 |
elif "\n\n" in answer:
|
493 |
possible_final_answer = answer.split("\n\n")[-1]
|
494 |
else:
|
495 |
possible_final_answer = answer.strip()
|
496 |
+
if len(possible_final_answer) == 1 and possible_final_answer in ['A', 'B', 'C', 'D', 'E']:
|
497 |
+
return possible_final_answer
|
498 |
+
elif len(possible_final_answer) > 1 and possible_final_answer[1] == ':' and possible_final_answer[0] in ['A', 'B', 'C', 'D', 'E']:
|
499 |
+
return possible_final_answer[0]
|
500 |
+
|
501 |
+
conversation = self.set_system_prompt(
|
502 |
+
[], "Transform the agent's answer to a single letter: 'A', 'B', 'C', 'D'.")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
503 |
conversation.append({"role": "user", "content": message +
|
504 |
+
"\nAgent's answer: " + answer + "\nAnswer (must be a letter):"})
|
505 |
+
return self.llm_infer(
|
506 |
+
messages=conversation,
|
507 |
+
temperature=temperature,
|
508 |
+
tools=None,
|
509 |
+
max_new_tokens=max_new_tokens,
|
510 |
+
max_token=max_token)
|
511 |
|
512 |
def run_summary_agent(self, thought_calls: str,
|
513 |
function_response: str,
|
514 |
temperature: float,
|
515 |
max_new_tokens: int,
|
516 |
+
max_token: int):
|
517 |
+
logger.info("Summarizing tool result")
|
518 |
+
prompt = f"""Thought and function calls:
|
519 |
{thought_calls}
|
|
|
520 |
Function calls' responses:
|
521 |
\"\"\"
|
522 |
{function_response}
|
523 |
\"\"\"
|
524 |
+
Summarize the function calls' responses in one sentence with all necessary information.
|
525 |
+
"""
|
526 |
+
conversation = [{"role": "user", "content": prompt}]
|
527 |
+
output = self.llm_infer(
|
528 |
+
messages=conversation,
|
529 |
+
temperature=temperature,
|
530 |
+
tools=None,
|
531 |
+
max_new_tokens=max_new_tokens,
|
532 |
+
max_token=max_token)
|
|
|
|
|
|
|
|
|
|
|
|
|
533 |
if '[' in output:
|
534 |
output = output.split('[')[0]
|
535 |
return output
|
536 |
|
537 |
def function_result_summary(self, input_list, status, enable_summary):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
538 |
if 'tool_call_step' not in status:
|
539 |
status['tool_call_step'] = 0
|
|
|
540 |
for idx in range(len(input_list)):
|
541 |
+
pos_id = len(input_list) - idx - 1
|
542 |
+
if input_list[pos_id]['role'] == 'assistant' and 'tool_calls' in input_list[pos_id]:
|
|
|
|
|
|
|
543 |
break
|
544 |
|
545 |
+
status['step'] = status.get('step', 0) + 1
|
|
|
|
|
|
|
|
|
546 |
if not enable_summary:
|
547 |
return status
|
548 |
|
549 |
+
status['summarized_index'] = status.get('summarized_index', 0)
|
550 |
+
status['summarized_step'] = status.get('summarized_step', 0)
|
551 |
+
status['previous_length'] = status.get('previous_length', 0)
|
552 |
+
status['history'] = status.get('history', [])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
553 |
|
554 |
function_response = ''
|
555 |
+
idx = status['summarized_index']
|
556 |
+
this_thought_calls = None
|
|
|
|
|
|
|
557 |
|
|
|
558 |
while idx < len(input_list):
|
559 |
+
if (self.summary_mode == 'step' and status['summarized_step'] < status['step'] - status['tool_call_step'] - self.summary_skip_last_k) or \
|
560 |
+
(self.summary_mode == 'length' and status['previous_length'] > self.summary_context_length):
|
561 |
if input_list[idx]['role'] == 'assistant':
|
562 |
+
if function_response:
|
563 |
+
status['summarized_step'] += 1
|
564 |
+
result_summary = self.run_summary_agent(
|
565 |
+
thought_calls=this_thought_calls,
|
566 |
+
function_response=function_response,
|
567 |
+
temperature=0.1,
|
568 |
+
max_new_tokens=512,
|
569 |
+
max_token=2048)
|
570 |
+
input_list.insert(last_call_idx + 1, {'role': 'tool', 'content': result_summary})
|
571 |
+
status['summarized_index'] = last_call_idx + 2
|
572 |
+
idx += 1
|
573 |
+
last_call_idx = idx
|
574 |
+
this_thought_calls = input_list[idx]['content'] + input_list[idx]['tool_calls']
|
575 |
+
function_response = ''
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
576 |
elif input_list[idx]['role'] == 'tool' and this_thought_calls is not None:
|
577 |
function_response += input_list[idx]['content']
|
578 |
del input_list[idx]
|
579 |
idx -= 1
|
|
|
580 |
else:
|
581 |
break
|
582 |
idx += 1
|
583 |
|
584 |
+
if function_response:
|
585 |
status['summarized_step'] += 1
|
586 |
result_summary = self.run_summary_agent(
|
587 |
thought_calls=this_thought_calls,
|
588 |
function_response=function_response,
|
589 |
temperature=0.1,
|
590 |
+
max_new_tokens=512,
|
591 |
+
max_token=2048)
|
|
|
|
|
592 |
tool_calls = json.loads(input_list[last_call_idx]['tool_calls'])
|
593 |
for tool_call in tool_calls:
|
594 |
del tool_call['call_id']
|
595 |
input_list[last_call_idx]['tool_calls'] = json.dumps(tool_calls)
|
596 |
+
input_list.insert(last_call_idx + 1, {'role': 'tool', 'content': result_summary})
|
|
|
597 |
status['summarized_index'] = last_call_idx + 2
|
598 |
|
599 |
return status
|
600 |
|
|
|
|
|
|
|
601 |
def update_parameters(self, **kwargs):
|
602 |
+
updated_attributes = {}
|
603 |
for key, value in kwargs.items():
|
604 |
if hasattr(self, key):
|
605 |
setattr(self, key, value)
|
606 |
+
updated_attributes[key] = value
|
607 |
+
logger.info("Updated parameters: %s", updated_attributes)
|
|
|
|
|
608 |
return updated_attributes
|
609 |
|
610 |
def run_gradio_chat(self, message: str,
|
|
|
614 |
max_token: int,
|
615 |
call_agent: bool,
|
616 |
conversation: gr.State,
|
617 |
+
max_round: int = 5,
|
618 |
seed: int = None,
|
619 |
call_agent_level: int = 0,
|
620 |
+
sub_agent_task: str = None,
|
621 |
+
uploaded_files: list = None):
|
622 |
+
logger.info("Chat started, message: %s", message[:100])
|
623 |
+
if not message or len(message.strip()) < 5:
|
624 |
+
yield "Please provide a valid message or upload files to analyze."
|
625 |
+
return
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
626 |
|
627 |
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
628 |
+
call_agent, call_agent_level, message)
|
|
|
|
|
|
|
629 |
conversation = self.initialize_conversation(
|
630 |
+
message, conversation, history)
|
|
|
|
|
631 |
history = []
|
632 |
+
last_outputs = []
|
633 |
|
634 |
next_round = True
|
|
|
635 |
current_round = 0
|
636 |
enable_summary = False
|
637 |
+
last_status = {}
|
638 |
token_overflow = False
|
|
|
|
|
|
|
639 |
|
640 |
try:
|
641 |
while next_round and current_round < max_round:
|
642 |
current_round += 1
|
643 |
+
if last_outputs:
|
644 |
function_call_messages, picked_tools_prompt, special_tool_call, current_gradio_history = yield from self.run_function_call_stream(
|
645 |
last_outputs, return_message=True,
|
646 |
existing_tools_prompt=picked_tools_prompt,
|
|
|
649 |
call_agent_level=call_agent_level,
|
650 |
temperature=temperature)
|
651 |
history.extend(current_gradio_history)
|
652 |
+
|
653 |
if special_tool_call == 'Finish':
|
654 |
yield history
|
655 |
next_round = False
|
656 |
conversation.extend(function_call_messages)
|
657 |
return function_call_messages[0]['content']
|
658 |
+
|
659 |
+
elif special_tool_call in ['RequireClarification', 'DirectResponse']:
|
660 |
+
last_msg = history[-1] if history else ChatMessage(role="assistant", content="Response needed.")
|
661 |
+
history.append(ChatMessage(role="assistant", content=last_msg.content))
|
662 |
yield history
|
663 |
next_round = False
|
664 |
+
return last_msg.content
|
665 |
+
|
666 |
if (self.enable_summary or token_overflow) and not call_agent:
|
|
|
|
|
667 |
enable_summary = True
|
668 |
last_status = self.function_result_summary(
|
669 |
+
conversation, status=last_status, enable_summary=enable_summary)
|
670 |
+
|
671 |
+
if function_call_messages:
|
672 |
conversation.extend(function_call_messages)
|
|
|
|
|
673 |
yield history
|
674 |
else:
|
675 |
next_round = False
|
676 |
+
conversation.append({"role": "assistant", "content": ''.join(last_outputs)})
|
|
|
677 |
return ''.join(last_outputs).replace("</s>", "")
|
678 |
+
|
|
|
|
|
|
|
|
|
|
|
679 |
last_outputs = []
|
680 |
last_outputs_str, token_overflow = self.llm_infer(
|
681 |
messages=conversation,
|
|
|
686 |
max_token=max_token,
|
687 |
seed=seed,
|
688 |
check_token_status=True)
|
689 |
+
|
690 |
+
if last_outputs_str is None:
|
691 |
+
logger.warning("Token limit exceeded")
|
692 |
+
if self.force_finish:
|
693 |
+
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
694 |
+
conversation, temperature, max_new_tokens, max_token)
|
695 |
+
history.append(ChatMessage(role="assistant", content=last_outputs_str.strip()))
|
696 |
+
yield history
|
697 |
+
return last_outputs_str
|
698 |
+
error_msg = "Token limit exceeded."
|
699 |
+
history.append(ChatMessage(role="assistant", content=error_msg))
|
700 |
+
yield history
|
701 |
+
return error_msg
|
702 |
+
|
703 |
last_thought = last_outputs_str.split("[TOOL_CALLS]")[0]
|
704 |
+
for msg in history:
|
705 |
+
if msg.metadata is not None:
|
706 |
+
msg.metadata['status'] = 'done'
|
707 |
+
|
708 |
if '[FinalAnswer]' in last_thought:
|
709 |
+
parts = last_thought.split('[FinalAnswer]', 1)
|
710 |
+
final_thought, final_answer = parts if len(parts) == 2 else (last_thought, "")
|
711 |
+
history.append(ChatMessage(role="assistant", content=final_thought.strip()))
|
|
|
|
|
|
|
712 |
yield history
|
713 |
+
history.append(ChatMessage(role="assistant", content="**🧠 Final Analysis:**\n" + final_answer.strip()))
|
|
|
|
|
|
|
714 |
yield history
|
715 |
else:
|
716 |
+
history.append(ChatMessage(role="assistant", content=last_thought))
|
|
|
717 |
yield history
|
718 |
|
719 |
last_outputs.append(last_outputs_str)
|
|
|
722 |
if self.force_finish:
|
723 |
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
724 |
conversation, temperature, max_new_tokens, max_token)
|
725 |
+
parts = last_outputs_str.split('[FinalAnswer]', 1)
|
726 |
+
final_thought, final_answer = parts if len(parts) == 2 else (last_outputs_str, "")
|
727 |
+
history.append(ChatMessage(role="assistant", content=final_thought.strip()))
|
728 |
+
yield history
|
729 |
+
history.append(ChatMessage(role="assistant", content="**🧠 Final Analysis:**\n" + final_answer.strip()))
|
730 |
+
yield history
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
731 |
else:
|
732 |
+
yield "Reasoning rounds exceeded limit."
|
733 |
|
734 |
except Exception as e:
|
735 |
+
logger.error("Exception in run_gradio_chat: %s", e, exc_info=True)
|
736 |
+
error_msg = f"Error: {e}"
|
737 |
+
history.append(ChatMessage(role="assistant", content=error_msg))
|
738 |
+
yield history
|
739 |
if self.force_finish:
|
740 |
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
741 |
+
conversation, temperature, max_new_tokens, max_token)
|
742 |
+
parts = last_outputs_str.split('[FinalAnswer]', 1)
|
743 |
+
final_thought, final_answer = parts if len(parts) == 2 else (last_outputs_str, "")
|
744 |
+
history.append(ChatMessage(role="assistant", content=final_thought.strip()))
|
745 |
+
yield history
|
746 |
+
history.append(ChatMessage(role="assistant", content="**🧠 Final Analysis:**\n" + final_answer.strip()))
|
747 |
+
yield history
|
748 |
+
return error_msg
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|