Update src/txagent/txagent.py
Browse files- src/txagent/txagent.py +397 -579
src/txagent/txagent.py
CHANGED
@@ -1,4 +1,3 @@
|
|
1 |
-
import gradio as gr
|
2 |
import os
|
3 |
import sys
|
4 |
import json
|
@@ -6,13 +5,13 @@ import gc
|
|
6 |
import numpy as np
|
7 |
from vllm import LLM, SamplingParams
|
8 |
from jinja2 import Template
|
9 |
-
from typing import List
|
10 |
import types
|
11 |
from tooluniverse import ToolUniverse
|
12 |
-
from gradio import ChatMessage
|
13 |
from .toolrag import ToolRAGModel
|
14 |
import torch
|
15 |
import logging
|
|
|
16 |
|
17 |
# Configure logging with a more specific logger name
|
18 |
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
|
@@ -21,28 +20,50 @@ logger = logging.getLogger("TxAgent")
|
|
21 |
from .utils import NoRepeatSentenceProcessor, ReasoningTraceChecker, tool_result_format
|
22 |
|
23 |
class TxAgent:
|
24 |
-
def __init__(self,
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
self.model_name = model_name
|
42 |
self.tokenizer = None
|
43 |
-
self.terminators = None
|
44 |
self.rag_model_name = rag_model_name
|
45 |
-
self.tool_files_dict = tool_files_dict
|
46 |
self.model = None
|
47 |
self.rag_model = ToolRAGModel(rag_model_name)
|
48 |
self.tooluniverse = None
|
@@ -61,106 +82,166 @@ class TxAgent:
|
|
61 |
self.avoid_repeat = avoid_repeat
|
62 |
self.seed = seed
|
63 |
self.enable_checker = enable_checker
|
64 |
-
self.additional_default_tools = additional_default_tools
|
65 |
logger.info("TxAgent initialized with model: %s, RAG: %s", model_name, rag_model_name)
|
66 |
|
67 |
-
def init_model(self):
|
|
|
68 |
self.load_models()
|
69 |
self.load_tooluniverse()
|
|
|
70 |
|
71 |
-
def load_models(self, model_name=None):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
if model_name is not None:
|
73 |
if model_name == self.model_name:
|
74 |
return f"The model {model_name} is already loaded."
|
75 |
self.model_name = model_name
|
76 |
|
77 |
-
|
78 |
-
model=
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
|
|
|
|
|
|
|
|
93 |
|
94 |
-
def load_tooluniverse(self):
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
self.tooluniverse.
|
99 |
-
|
100 |
-
|
|
|
|
|
|
|
|
|
|
|
101 |
|
102 |
-
def load_tool_desc_embedding(self):
|
|
|
103 |
cache_path = os.path.join(os.path.dirname(self.tool_files_dict["new_tool"]), "tool_embeddings.pkl")
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
|
|
|
|
|
|
|
|
110 |
|
111 |
-
def rag_infer(self, query, top_k=5):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
112 |
return self.rag_model.rag_infer(query, top_k)
|
113 |
|
114 |
-
def
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
|
|
|
|
|
|
|
|
125 |
if conversation is None:
|
126 |
conversation = []
|
127 |
|
128 |
-
conversation = self.set_system_prompt(
|
129 |
-
conversation, self.prompt_multi_step)
|
130 |
if history:
|
131 |
-
for
|
132 |
-
if
|
133 |
-
conversation.append({"role": "user", "content":
|
134 |
-
elif
|
135 |
-
conversation.append({"role": "assistant", "content":
|
136 |
conversation.append({"role": "user", "content": message})
|
137 |
logger.debug("Conversation initialized with %d messages", len(conversation))
|
138 |
return conversation
|
139 |
|
140 |
-
def tool_RAG(self,
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
145 |
if not self.enable_rag:
|
146 |
-
return []
|
|
|
147 |
extra_factor = 10
|
148 |
if picked_tool_names is None:
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
picked_tool_names_no_special = [
|
|
|
|
|
|
|
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(picked_tools)
|
158 |
logger.debug("Retrieved %d tools via RAG", len(picked_tools_prompt))
|
|
|
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('Finish', return_prompt=True))
|
166 |
logger.debug("Finish tool added")
|
@@ -169,25 +250,37 @@ class TxAgent:
|
|
169 |
logger.debug("CallAgent tool added")
|
170 |
return tools
|
171 |
|
172 |
-
def
|
173 |
-
|
174 |
-
logger.debug("Finish tool added")
|
175 |
-
return tools
|
176 |
-
|
177 |
-
def set_system_prompt(self, conversation, sys_prompt):
|
178 |
if not conversation:
|
179 |
conversation.append({"role": "system", "content": sys_prompt})
|
180 |
else:
|
181 |
conversation[0] = {"role": "system", "content": sys_prompt}
|
182 |
return conversation
|
183 |
|
184 |
-
def run_function_call(self,
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
191 |
try:
|
192 |
function_call_json, message = self.tooluniverse.extract_function_call_json(
|
193 |
fcall_str, return_message=return_message, verbose=False)
|
@@ -206,13 +299,12 @@ class TxAgent:
|
|
206 |
special_tool_call = 'Finish'
|
207 |
break
|
208 |
elif function_call_json[i]["name"] == 'CallAgent':
|
209 |
-
if call_agent_level < 2 and call_agent:
|
210 |
solution_plan = function_call_json[i]['arguments']['solution']
|
211 |
full_message = (
|
212 |
-
message_for_call_agent +
|
213 |
"\nYou must follow the following plan to answer the question: " +
|
214 |
str(solution_plan)
|
215 |
-
)
|
216 |
call_result = self.run_multistep_agent(
|
217 |
full_message, temperature=temperature,
|
218 |
max_new_tokens=512, max_token=131072,
|
@@ -230,7 +322,11 @@ class TxAgent:
|
|
230 |
logger.info("Tool Call Result: %s", call_result)
|
231 |
call_results.append({
|
232 |
"role": "tool",
|
233 |
-
"content": json.dumps({
|
|
|
|
|
|
|
|
|
234 |
})
|
235 |
else:
|
236 |
call_results.append({
|
@@ -243,109 +339,99 @@ class TxAgent:
|
|
243 |
"content": message.strip(),
|
244 |
"tool_calls": json.dumps(function_call_json)
|
245 |
}] + call_results
|
246 |
-
return revised_messages, existing_tools_prompt, special_tool_call
|
247 |
-
|
248 |
-
def
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
|
255 |
-
|
256 |
-
|
257 |
-
|
258 |
-
|
259 |
-
|
260 |
-
|
261 |
-
|
262 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
263 |
|
264 |
-
|
265 |
-
|
266 |
-
|
267 |
-
|
268 |
-
|
269 |
-
if isinstance(function_call_json, list):
|
270 |
-
for i in range(len(function_call_json)):
|
271 |
-
if function_call_json[i]["name"] == 'Finish':
|
272 |
-
special_tool_call = 'Finish'
|
273 |
-
break
|
274 |
-
elif function_call_json[i]["name"] == 'DirectResponse':
|
275 |
-
call_result = function_call_json[i]['arguments']['respose']
|
276 |
-
special_tool_call = 'DirectResponse'
|
277 |
-
elif function_call_json[i]["name"] == 'RequireClarification':
|
278 |
-
call_result = function_call_json[i]['arguments']['unclear_question']
|
279 |
-
special_tool_call = 'RequireClarification'
|
280 |
-
elif function_call_json[i]["name"] == 'CallAgent':
|
281 |
-
if call_agent_level < 2 and call_agent:
|
282 |
-
solution_plan = function_call_json[i]['arguments']['solution']
|
283 |
-
full_message = (
|
284 |
-
message_for_call_agent +
|
285 |
-
"\nYou must follow the following plan to answer the question: " +
|
286 |
-
str(solution_plan)
|
287 |
-
)
|
288 |
-
sub_agent_task = "Sub TxAgent plan: " + str(solution_plan)
|
289 |
-
call_result = yield from self.run_gradio_chat(
|
290 |
-
full_message, history=[], temperature=temperature,
|
291 |
-
max_new_tokens=512, max_token=131072,
|
292 |
-
call_agent=False, call_agent_level=call_agent_level,
|
293 |
-
conversation=None, sub_agent_task=sub_agent_task)
|
294 |
-
if call_result is not None and isinstance(call_result, str):
|
295 |
-
call_result = call_result.split('[FinalAnswer]')[-1]
|
296 |
-
else:
|
297 |
-
call_result = "⚠️ No content returned from sub-agent."
|
298 |
-
else:
|
299 |
-
call_result = "Error: CallAgent disabled."
|
300 |
-
else:
|
301 |
-
call_result = self.tooluniverse.run_one_function(function_call_json[i])
|
302 |
-
call_id = self.tooluniverse.call_id_gen()
|
303 |
-
function_call_json[i]["call_id"] = call_id
|
304 |
-
call_results.append({
|
305 |
-
"role": "tool",
|
306 |
-
"content": json.dumps({"tool_name": function_call_json[i]["name"], "content": call_result, "call_id": call_id})
|
307 |
-
})
|
308 |
-
if return_gradio_history and function_call_json[i]["name"] != 'Finish':
|
309 |
-
metadata = {"title": f"🧰 {function_call_json[i]['name']}", "log": str(function_call_json[i]['arguments'])}
|
310 |
-
gradio_history.append(ChatMessage(role="assistant", content=str(call_result), metadata=metadata))
|
311 |
-
else:
|
312 |
-
call_results.append({
|
313 |
-
"role": "tool",
|
314 |
-
"content": json.dumps({"content": "Invalid or no function call detected."})
|
315 |
-
})
|
316 |
|
317 |
-
|
318 |
-
|
319 |
-
|
320 |
-
|
321 |
-
}] + call_results
|
322 |
-
if return_gradio_history:
|
323 |
-
return revised_messages, existing_tools_prompt, special_tool_call, gradio_history
|
324 |
-
return revised_messages, existing_tools_prompt, special_tool_call
|
325 |
|
326 |
-
|
327 |
-
if
|
328 |
-
|
329 |
-
|
330 |
-
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
-
|
335 |
-
output_begin_string='[FinalAnswer]',
|
336 |
-
skip_special_tokens=True,
|
337 |
-
max_new_tokens=max_new_tokens,
|
338 |
-
max_token=max_token)
|
339 |
-
logger.info("Unfinished reasoning answer: %s", last_outputs_str[:100])
|
340 |
-
return last_outputs_str
|
341 |
|
342 |
-
|
343 |
-
|
344 |
-
|
345 |
-
|
346 |
-
|
347 |
-
|
348 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
349 |
logger.info("Starting multistep agent for message: %s", message[:100])
|
350 |
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
351 |
call_agent, call_agent_level, message)
|
@@ -415,196 +501,109 @@ class TxAgent:
|
|
415 |
conversation, temperature, max_new_tokens, max_token)
|
416 |
return None
|
417 |
|
418 |
-
def
|
419 |
-
|
420 |
-
|
421 |
-
|
422 |
-
|
423 |
-
|
424 |
-
|
425 |
-
|
426 |
-
|
427 |
-
|
428 |
-
|
429 |
-
|
430 |
-
|
431 |
-
|
432 |
-
|
433 |
-
|
434 |
-
|
435 |
-
)
|
436 |
-
|
437 |
-
|
438 |
-
|
439 |
-
|
440 |
-
|
441 |
-
|
442 |
-
if check_token_status and max_token is not None:
|
443 |
-
token_overflow = False
|
444 |
-
num_input_tokens = len(self.tokenizer.encode(prompt, add_special_tokens=False))
|
445 |
-
logger.info("Input prompt tokens: %d, max_token: %d", num_input_tokens, max_token)
|
446 |
-
if num_input_tokens > max_token:
|
447 |
-
torch.cuda.empty_cache()
|
448 |
-
gc.collect()
|
449 |
-
logger.warning("Token overflow: %d > %d", num_input_tokens, max_token)
|
450 |
-
return None, True
|
451 |
|
452 |
-
|
453 |
-
|
454 |
-
|
455 |
-
|
456 |
-
|
457 |
-
|
458 |
-
|
459 |
-
|
460 |
-
|
461 |
-
|
462 |
-
|
463 |
-
|
464 |
-
|
465 |
-
|
466 |
-
logger.info("Starting self agent")
|
467 |
-
conversation = self.set_system_prompt([], self.self_prompt)
|
468 |
-
conversation.append({"role": "user", "content": message})
|
469 |
-
return self.llm_infer(
|
470 |
-
messages=conversation,
|
471 |
-
temperature=temperature,
|
472 |
-
tools=None,
|
473 |
-
max_new_tokens=max_new_tokens,
|
474 |
-
max_token=max_token)
|
475 |
|
476 |
-
|
477 |
-
|
478 |
-
|
479 |
-
max_token: int):
|
480 |
-
logger.info("Starting chat agent")
|
481 |
-
conversation = self.set_system_prompt([], self.chat_prompt)
|
482 |
-
conversation.append({"role": "user", "content": message})
|
483 |
-
return self.llm_infer(
|
484 |
-
messages=conversation,
|
485 |
-
temperature=temperature,
|
486 |
-
tools=None,
|
487 |
-
max_new_tokens=max_new_tokens,
|
488 |
-
max_token=max_token)
|
489 |
|
490 |
-
|
491 |
-
|
492 |
-
|
493 |
-
|
494 |
-
|
495 |
-
|
496 |
-
|
497 |
-
|
498 |
-
|
499 |
-
|
500 |
-
|
501 |
-
|
502 |
-
|
503 |
-
|
504 |
-
|
505 |
-
|
506 |
-
|
507 |
-
|
508 |
-
|
509 |
-
|
510 |
-
|
511 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
512 |
messages=conversation,
|
513 |
temperature=temperature,
|
514 |
-
tools=
|
|
|
|
|
515 |
max_new_tokens=max_new_tokens,
|
516 |
max_token=max_token)
|
|
|
|
|
517 |
|
518 |
-
def
|
519 |
-
|
520 |
-
|
521 |
-
|
522 |
-
|
523 |
-
|
524 |
-
|
525 |
-
|
526 |
-
|
527 |
-
|
528 |
-
{function_response}
|
529 |
-
\"\"\"
|
530 |
-
Summarize the function calls' l responses in one sentence with all necessary information.
|
531 |
-
"""
|
532 |
-
conversation = [{"role": "user", "content": prompt}]
|
533 |
-
output = self.llm_infer(
|
534 |
-
messages=conversation,
|
535 |
-
temperature=temperature,
|
536 |
-
tools=None,
|
537 |
-
max_new_tokens=max_new_tokens,
|
538 |
-
max_token=max_token)
|
539 |
-
if '[' in output:
|
540 |
-
output = output.split('[')[0]
|
541 |
-
return output
|
542 |
-
|
543 |
-
def function_result_summary(self, input_list, status, enable_summary):
|
544 |
-
if 'tool_call_step' not in status:
|
545 |
-
status['tool_call_step'] = 0
|
546 |
-
for idx in range(len(input_list)):
|
547 |
-
pos_id = len(input_list) - idx - 1
|
548 |
-
if input_list[pos_id]['role'] == 'assistant' and 'tool_calls' in input_list[pos_id]:
|
549 |
-
break
|
550 |
-
|
551 |
-
status['step'] = status.get('step', 0) + 1
|
552 |
-
if not enable_summary:
|
553 |
-
return status
|
554 |
-
|
555 |
-
status['summarized_index'] = status.get('summarized_index', 0)
|
556 |
-
status['summarized_step'] = status.get('summarized_step', 0)
|
557 |
-
status['previous_length'] = status.get('previous_length', 0)
|
558 |
-
status['history'] = status.get('history', [])
|
559 |
-
|
560 |
-
function_response = ''
|
561 |
-
idx = status['summarized_index']
|
562 |
-
this_thought_calls = None
|
563 |
-
|
564 |
-
while idx < len(input_list):
|
565 |
-
if (self.summary_mode == 'step' and status['summarized_step'] < status['step'] - status['tool_call_step'] - self.summary_skip_last_k) or \
|
566 |
-
(self.summary_mode == 'length' and status['previous_length'] > self.summary_context_length):
|
567 |
-
if input_list[idx]['role'] == 'assistant':
|
568 |
-
if function_response:
|
569 |
-
status['summarized_step'] += 1
|
570 |
-
result_summary = self.run_summary_agent(
|
571 |
-
thought_calls=this_thought_calls,
|
572 |
-
function_response=function_response,
|
573 |
-
temperature=0.1,
|
574 |
-
max_new_tokens=512,
|
575 |
-
max_token=131072)
|
576 |
-
input_list.insert(last_call_idx + 1, {'role': 'tool', 'content': result_summary})
|
577 |
-
status['summarized_index'] = last_call_idx + 2
|
578 |
-
idx += 1
|
579 |
-
last_call_idx = idx
|
580 |
-
this_thought_calls = input_list[idx]['content'] + input_list[idx]['tool_calls']
|
581 |
-
function_response = ''
|
582 |
-
elif input_list[idx]['role'] == 'tool' and this_thought_calls is not None:
|
583 |
-
function_response += input_list[idx]['content']
|
584 |
-
del input_list[idx]
|
585 |
-
idx -= 1
|
586 |
-
else:
|
587 |
-
break
|
588 |
-
idx += 1
|
589 |
-
|
590 |
-
if function_response:
|
591 |
-
status['summarized_step'] += 1
|
592 |
-
result_summary = self.run_summary_agent(
|
593 |
-
thought_calls=this_thought_calls,
|
594 |
-
function_response=function_response,
|
595 |
-
temperature=0.1,
|
596 |
-
max_new_tokens=512,
|
597 |
-
max_token=131072)
|
598 |
-
tool_calls = json.loads(input_list[last_call_idx]['tool_calls'])
|
599 |
-
for tool_call in tool_calls:
|
600 |
-
del tool_call['call_id']
|
601 |
-
input_list[last_call_idx]['tool_calls'] = json.dumps(tool_calls)
|
602 |
-
input_list.insert(last_call_idx + 1, {'role': 'tool', 'content': result_summary})
|
603 |
-
status['summarized_index'] = last_call_idx + 2
|
604 |
-
|
605 |
-
return status
|
606 |
-
|
607 |
-
def update_parameters(self, **kwargs):
|
608 |
updated_attributes = {}
|
609 |
for key, value in kwargs.items():
|
610 |
if hasattr(self, key):
|
@@ -613,199 +612,18 @@ Summarize the function calls' l responses in one sentence with all necessary inf
|
|
613 |
logger.info("Updated parameters: %s", updated_attributes)
|
614 |
return updated_attributes
|
615 |
|
616 |
-
def
|
617 |
-
|
618 |
-
|
619 |
-
|
620 |
-
|
621 |
-
|
622 |
-
|
623 |
-
|
624 |
-
|
625 |
-
|
626 |
-
|
627 |
-
uploaded_files: list = None):
|
628 |
-
logger.info("Chat started, message: %s", message[:100])
|
629 |
-
if not message or len(message.strip()) < 5:
|
630 |
-
yield "Please provide a valid message or upload files to analyze."
|
631 |
-
return
|
632 |
-
|
633 |
-
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
634 |
-
call_agent, call_agent_level, message)
|
635 |
-
conversation = self.initialize_conversation(
|
636 |
-
message, conversation, history)
|
637 |
-
history = []
|
638 |
-
last_outputs = []
|
639 |
-
|
640 |
-
next_round = True
|
641 |
-
current_round = 0
|
642 |
-
enable_summary = False
|
643 |
-
last_status = {}
|
644 |
-
token_overflow = False
|
645 |
-
|
646 |
-
try:
|
647 |
-
while next_round and current_round < max_round:
|
648 |
-
current_round += 1
|
649 |
-
logger.debug("Starting round %d/%d", current_round, max_round)
|
650 |
-
if last_outputs:
|
651 |
-
function_call_messages, picked_tools_prompt, special_tool_call, current_gradio_history = yield from self.run_function_call_stream(
|
652 |
-
last_outputs, return_message=True,
|
653 |
-
existing_tools_prompt=picked_tools_prompt,
|
654 |
-
message_for_call_agent=message,
|
655 |
-
call_agent=call_agent,
|
656 |
-
call_agent_level=call_agent_level,
|
657 |
-
temperature=temperature)
|
658 |
-
history.extend(current_gradio_history)
|
659 |
-
|
660 |
-
if special_tool_call == 'Finish':
|
661 |
-
logger.info("Finish tool called, ending chat")
|
662 |
-
yield history
|
663 |
-
next_round = False
|
664 |
-
conversation.extend(function_call_messages)
|
665 |
-
content = function_call_messages[0]['content']
|
666 |
-
if content:
|
667 |
-
return content
|
668 |
-
return "No content returned after Finish tool call."
|
669 |
-
|
670 |
-
elif special_tool_call in ['RequireClarification', 'DirectResponse']:
|
671 |
-
last_msg = history[-1] if history else ChatMessage(role="assistant", content="Response needed.")
|
672 |
-
history.append(ChatMessage(role="assistant", content=last_msg.content))
|
673 |
-
logger.info("Special tool %s called, ending chat", special_tool_call)
|
674 |
-
yield history
|
675 |
-
next_round = False
|
676 |
-
return last_msg.content
|
677 |
-
|
678 |
-
if (self.enable_summary or token_overflow) and not call_agent:
|
679 |
-
enable_summary = True
|
680 |
-
last_status = self.function_result_summary(
|
681 |
-
conversation, status=last_status, enable_summary=enable_summary)
|
682 |
-
|
683 |
-
if function_call_messages:
|
684 |
-
conversation.extend(function_call_messages)
|
685 |
-
yield history
|
686 |
-
else:
|
687 |
-
next_round = False
|
688 |
-
conversation.append({"role": "assistant", "content": ''.join(last_outputs)})
|
689 |
-
logger.info("No function call messages, ending chat")
|
690 |
-
return ''.join(last_outputs).replace("</s>", "")
|
691 |
-
|
692 |
-
last_outputs = []
|
693 |
-
last_outputs_str, token_overflow = self.llm_infer(
|
694 |
-
messages=conversation,
|
695 |
-
temperature=temperature,
|
696 |
-
tools=picked_tools_prompt,
|
697 |
-
skip_special_tokens=False,
|
698 |
-
max_new_tokens=max_new_tokens,
|
699 |
-
max_token=max_token,
|
700 |
-
seed=seed,
|
701 |
-
check_token_status=True)
|
702 |
-
|
703 |
-
if last_outputs_str is None:
|
704 |
-
logger.warning("Token limit exceeded")
|
705 |
-
if self.force_finish:
|
706 |
-
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
707 |
-
conversation, temperature, max_new_tokens, max_token)
|
708 |
-
history.append(ChatMessage(role="assistant", content=last_outputs_str.strip()))
|
709 |
-
yield history
|
710 |
-
return last_outputs_str
|
711 |
-
error_msg = "Token limit exceeded."
|
712 |
-
history.append(ChatMessage(role="assistant", content=error_msg))
|
713 |
-
yield history
|
714 |
-
return error_msg
|
715 |
-
|
716 |
-
last_thought = last_outputs_str.split("[TOOL_CALLS]")[0]
|
717 |
-
for msg in history:
|
718 |
-
if msg.metadata is not None:
|
719 |
-
msg.metadata['status'] = 'done'
|
720 |
-
|
721 |
-
if '[FinalAnswer]' in last_thought:
|
722 |
-
parts = last_thought.split('[FinalAnswer]', 1)
|
723 |
-
final_thought, final_answer = parts if len(parts) == 2 else (last_thought, "")
|
724 |
-
history.append(ChatMessage(role="assistant", content=final_thought.strip()))
|
725 |
-
yield history
|
726 |
-
history.append(ChatMessage(role="assistant", content="**🧠 Final Analysis:**\n" + final_answer.strip()))
|
727 |
-
logger.info("Final answer provided: %s", final_answer[:100])
|
728 |
-
yield history
|
729 |
-
next_round = False # Ensure we exit after final answer
|
730 |
-
return final_answer
|
731 |
-
else:
|
732 |
-
history.append(ChatMessage(role="assistant", content=last_thought))
|
733 |
-
yield history
|
734 |
-
|
735 |
-
last_outputs.append(last_outputs_str)
|
736 |
-
|
737 |
-
if next_round:
|
738 |
-
if self.force_finish:
|
739 |
-
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
740 |
-
conversation, temperature, max_new_tokens, max_token)
|
741 |
-
parts = last_outputs_str.split('[FinalAnswer]', 1)
|
742 |
-
final_thought, final_answer = parts if len(parts) == 2 else (last_outputs_str, "")
|
743 |
-
history.append(ChatMessage(role="assistant", content=final_thought.strip()))
|
744 |
-
yield history
|
745 |
-
history.append(ChatMessage(role="assistant", content="**🧠 Final Analysis:**\n" + final_answer.strip()))
|
746 |
-
logger.info("Forced final answer: %s", final_answer[:100])
|
747 |
-
yield history
|
748 |
-
return final_answer
|
749 |
-
else:
|
750 |
-
error_msg = "Reasoning rounds exceeded limit."
|
751 |
-
history.append(ChatMessage(role="assistant", content=error_msg))
|
752 |
-
yield history
|
753 |
-
return error_msg
|
754 |
|
755 |
-
|
756 |
-
|
757 |
-
|
758 |
-
history.append(ChatMessage(role="assistant", content=error_msg))
|
759 |
-
yield history
|
760 |
-
if self.force_finish:
|
761 |
-
last_outputs_str = self.get_answer_based_on_unfinished_reasoning(
|
762 |
-
conversation, temperature, max_new_tokens, max_token)
|
763 |
-
parts = last_outputs_str.split('[FinalAnswer]', 1)
|
764 |
-
final_thought, final_answer = parts if len(parts) == 2 else (last_outputs_str, "")
|
765 |
-
history.append(ChatMessage(role="assistant", content=final_thought.strip()))
|
766 |
-
yield history
|
767 |
-
history.append(ChatMessage(role="assistant", content="**🧠 Final Analysis:**\n" + final_answer.strip()))
|
768 |
-
logger.info("Forced final answer after error: %s", final_answer[:100])
|
769 |
-
yield history
|
770 |
-
return final_answer
|
771 |
-
return error_msg
|
772 |
-
|
773 |
-
def run_gradio_chat_batch(self, messages: List[str],
|
774 |
-
temperature: float,
|
775 |
-
max_new_tokens: int = 2048,
|
776 |
-
max_token: int = 131072,
|
777 |
-
call_agent: bool = False,
|
778 |
-
conversation: List = None,
|
779 |
-
max_round: int = 5,
|
780 |
-
seed: int = None,
|
781 |
-
call_agent_level: int = 0):
|
782 |
-
"""Run batch inference for multiple messages."""
|
783 |
-
logger.info("Starting batch chat for %d messages", len(messages))
|
784 |
-
batch_results = []
|
785 |
-
|
786 |
-
for message in messages:
|
787 |
-
# Initialize conversation for each message
|
788 |
-
conv = self.initialize_conversation(message, conversation, history=None)
|
789 |
-
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
790 |
-
call_agent, call_agent_level, message)
|
791 |
-
|
792 |
-
# Run single inference for simplicity (extend for multi-round if needed)
|
793 |
-
output, token_overflow = self.llm_infer(
|
794 |
-
messages=conv,
|
795 |
-
temperature=temperature,
|
796 |
-
tools=picked_tools_prompt,
|
797 |
-
max_new_tokens=max_new_tokens,
|
798 |
-
max_token=max_token,
|
799 |
-
skip_special_tokens=False,
|
800 |
-
seed=seed,
|
801 |
-
check_token_status=True
|
802 |
-
)
|
803 |
-
|
804 |
-
if output is None:
|
805 |
-
logger.warning("Token limit exceeded for message: %s", message[:100])
|
806 |
-
batch_results.append("Token limit exceeded.")
|
807 |
-
else:
|
808 |
-
batch_results.append(output)
|
809 |
-
|
810 |
-
logger.info("Batch chat completed for %d messages", len(messages))
|
811 |
-
return batch_results
|
|
|
|
|
1 |
import os
|
2 |
import sys
|
3 |
import json
|
|
|
5 |
import numpy as np
|
6 |
from vllm import LLM, SamplingParams
|
7 |
from jinja2 import Template
|
8 |
+
from typing import List, Dict, Optional, Union, Generator
|
9 |
import types
|
10 |
from tooluniverse import ToolUniverse
|
|
|
11 |
from .toolrag import ToolRAGModel
|
12 |
import torch
|
13 |
import logging
|
14 |
+
from datetime import datetime
|
15 |
|
16 |
# Configure logging with a more specific logger name
|
17 |
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
|
|
|
20 |
from .utils import NoRepeatSentenceProcessor, ReasoningTraceChecker, tool_result_format
|
21 |
|
22 |
class TxAgent:
|
23 |
+
def __init__(self,
|
24 |
+
model_name: str,
|
25 |
+
rag_model_name: str,
|
26 |
+
tool_files_dict: Optional[Dict] = None,
|
27 |
+
enable_finish: bool = True,
|
28 |
+
enable_rag: bool = False,
|
29 |
+
enable_summary: bool = False,
|
30 |
+
init_rag_num: int = 0,
|
31 |
+
step_rag_num: int = 0,
|
32 |
+
summary_mode: str = 'step',
|
33 |
+
summary_skip_last_k: int = 0,
|
34 |
+
summary_context_length: Optional[int] = None,
|
35 |
+
force_finish: bool = True,
|
36 |
+
avoid_repeat: bool = True,
|
37 |
+
seed: Optional[int] = None,
|
38 |
+
enable_checker: bool = False,
|
39 |
+
enable_chat: bool = False,
|
40 |
+
additional_default_tools: Optional[List] = None):
|
41 |
+
"""
|
42 |
+
Initialize the TxAgent with specified configuration.
|
43 |
+
|
44 |
+
Args:
|
45 |
+
model_name: Name of the main LLM model
|
46 |
+
rag_model_name: Name of the RAG model
|
47 |
+
tool_files_dict: Dictionary of tool files
|
48 |
+
enable_finish: Whether to enable the Finish tool
|
49 |
+
enable_rag: Whether to enable RAG functionality
|
50 |
+
enable_summary: Whether to enable summarization
|
51 |
+
init_rag_num: Initial number of RAG tools to retrieve
|
52 |
+
step_rag_num: Number of RAG tools to retrieve per step
|
53 |
+
summary_mode: Mode for summarization ('step' or 'length')
|
54 |
+
summary_skip_last_k: Number of last steps to skip in summarization
|
55 |
+
summary_context_length: Context length threshold for summarization
|
56 |
+
force_finish: Whether to force finish when max rounds reached
|
57 |
+
avoid_repeat: Whether to avoid repeating similar responses
|
58 |
+
seed: Random seed for reproducibility
|
59 |
+
enable_checker: Whether to enable reasoning trace checker
|
60 |
+
enable_chat: Whether to enable chat mode
|
61 |
+
additional_default_tools: Additional tools to include by default
|
62 |
+
"""
|
63 |
self.model_name = model_name
|
64 |
self.tokenizer = None
|
|
|
65 |
self.rag_model_name = rag_model_name
|
66 |
+
self.tool_files_dict = tool_files_dict or {}
|
67 |
self.model = None
|
68 |
self.rag_model = ToolRAGModel(rag_model_name)
|
69 |
self.tooluniverse = None
|
|
|
82 |
self.avoid_repeat = avoid_repeat
|
83 |
self.seed = seed
|
84 |
self.enable_checker = enable_checker
|
85 |
+
self.additional_default_tools = additional_default_tools or []
|
86 |
logger.info("TxAgent initialized with model: %s, RAG: %s", model_name, rag_model_name)
|
87 |
|
88 |
+
def init_model(self) -> None:
|
89 |
+
"""Initialize both the main model and tool universe."""
|
90 |
self.load_models()
|
91 |
self.load_tooluniverse()
|
92 |
+
logger.info("Model and tools initialized successfully")
|
93 |
|
94 |
+
def load_models(self, model_name: Optional[str] = None) -> str:
|
95 |
+
"""
|
96 |
+
Load the specified model or the default model if none specified.
|
97 |
+
|
98 |
+
Args:
|
99 |
+
model_name: Name of the model to load
|
100 |
+
|
101 |
+
Returns:
|
102 |
+
Status message indicating success or failure
|
103 |
+
"""
|
104 |
if model_name is not None:
|
105 |
if model_name == self.model_name:
|
106 |
return f"The model {model_name} is already loaded."
|
107 |
self.model_name = model_name
|
108 |
|
109 |
+
try:
|
110 |
+
self.model = LLM(
|
111 |
+
model=self.model_name,
|
112 |
+
dtype="float16",
|
113 |
+
max_model_len=131072,
|
114 |
+
max_num_batched_tokens=65536,
|
115 |
+
max_num_seqs=512,
|
116 |
+
gpu_memory_utilization=0.95,
|
117 |
+
trust_remote_code=True,
|
118 |
+
)
|
119 |
+
self.tokenizer = self.model.get_tokenizer()
|
120 |
+
self.chat_template = Template(self.tokenizer.chat_template)
|
121 |
+
logger.info(
|
122 |
+
"Model %s loaded with max_model_len=%d, max_num_batched_tokens=%d",
|
123 |
+
self.model_name, 131072, 65536
|
124 |
+
)
|
125 |
+
return f"Model {model_name} loaded successfully."
|
126 |
+
except Exception as e:
|
127 |
+
logger.error("Failed to load model: %s", str(e))
|
128 |
+
raise RuntimeError(f"Failed to load model: {str(e)}")
|
129 |
|
130 |
+
def load_tooluniverse(self) -> None:
|
131 |
+
"""Load and initialize the tool universe with specified tools."""
|
132 |
+
try:
|
133 |
+
self.tooluniverse = ToolUniverse(tool_files=self.tool_files_dict)
|
134 |
+
self.tooluniverse.load_tools()
|
135 |
+
special_tools = self.tooluniverse.prepare_tool_prompts(
|
136 |
+
self.tooluniverse.tool_category_dicts["special_tools"])
|
137 |
+
self.special_tools_name = [tool['name'] for tool in special_tools]
|
138 |
+
logger.info("ToolUniverse loaded with %d special tools", len(self.special_tools_name))
|
139 |
+
except Exception as e:
|
140 |
+
logger.error("Failed to load tools: %s", str(e))
|
141 |
+
raise RuntimeError(f"Failed to load tools: {str(e)}")
|
142 |
|
143 |
+
def load_tool_desc_embedding(self) -> None:
|
144 |
+
"""Load tool description embeddings from cache or generate new ones."""
|
145 |
cache_path = os.path.join(os.path.dirname(self.tool_files_dict["new_tool"]), "tool_embeddings.pkl")
|
146 |
+
try:
|
147 |
+
if os.path.exists(cache_path):
|
148 |
+
self.rag_model.load_cached_embeddings(cache_path)
|
149 |
+
else:
|
150 |
+
self.rag_model.load_tool_desc_embedding(self.tooluniverse)
|
151 |
+
self.rag_model.save_embeddings(cache_path)
|
152 |
+
logger.info("Tool description embeddings loaded successfully")
|
153 |
+
except Exception as e:
|
154 |
+
logger.error("Failed to load tool embeddings: %s", str(e))
|
155 |
+
raise RuntimeError(f"Failed to load tool embeddings: {str(e)}")
|
156 |
|
157 |
+
def rag_infer(self, query: str, top_k: int = 5) -> List[str]:
|
158 |
+
"""
|
159 |
+
Perform RAG inference to retrieve relevant tools.
|
160 |
+
|
161 |
+
Args:
|
162 |
+
query: The query to search for
|
163 |
+
top_k: Number of top results to return
|
164 |
+
|
165 |
+
Returns:
|
166 |
+
List of relevant tool names
|
167 |
+
"""
|
168 |
+
if not self.enable_rag:
|
169 |
+
return []
|
170 |
return self.rag_model.rag_infer(query, top_k)
|
171 |
|
172 |
+
def initialize_conversation(self,
|
173 |
+
message: str,
|
174 |
+
conversation: Optional[List[Dict]] = None,
|
175 |
+
history: Optional[List[Dict]] = None) -> List[Dict]:
|
176 |
+
"""
|
177 |
+
Initialize or extend a conversation with the given message and history.
|
178 |
+
|
179 |
+
Args:
|
180 |
+
message: The new message to add
|
181 |
+
conversation: Existing conversation to extend
|
182 |
+
history: Chat history to incorporate
|
183 |
+
|
184 |
+
Returns:
|
185 |
+
Updated conversation list
|
186 |
+
"""
|
187 |
if conversation is None:
|
188 |
conversation = []
|
189 |
|
190 |
+
conversation = self.set_system_prompt(conversation, self.prompt_multi_step)
|
|
|
191 |
if history:
|
192 |
+
for msg in history:
|
193 |
+
if msg['role'] == 'user':
|
194 |
+
conversation.append({"role": "user", "content": msg['content']})
|
195 |
+
elif msg['role'] == 'assistant':
|
196 |
+
conversation.append({"role": "assistant", "content": msg['content']})
|
197 |
conversation.append({"role": "user", "content": message})
|
198 |
logger.debug("Conversation initialized with %d messages", len(conversation))
|
199 |
return conversation
|
200 |
|
201 |
+
def tool_RAG(self,
|
202 |
+
message: Optional[str] = None,
|
203 |
+
picked_tool_names: Optional[List[str]] = None,
|
204 |
+
existing_tools_prompt: List = [],
|
205 |
+
rag_num: int = 0,
|
206 |
+
return_call_result: bool = False) -> Union[List, Tuple[List, List]]:
|
207 |
+
"""
|
208 |
+
Retrieve relevant tools using RAG.
|
209 |
+
|
210 |
+
Args:
|
211 |
+
message: The query message for RAG
|
212 |
+
picked_tool_names: Pre-selected tool names
|
213 |
+
existing_tools_prompt: Existing tools to include
|
214 |
+
rag_num: Number of tools to retrieve
|
215 |
+
return_call_result: Whether to return tool names
|
216 |
+
|
217 |
+
Returns:
|
218 |
+
List of tool prompts or tuple with tool names if return_call_result is True
|
219 |
+
"""
|
220 |
if not self.enable_rag:
|
221 |
+
return [] if not return_call_result else ([], [])
|
222 |
+
|
223 |
extra_factor = 10
|
224 |
if picked_tool_names is None:
|
225 |
+
if message is None:
|
226 |
+
raise ValueError("Either message or picked_tool_names must be provided")
|
227 |
+
picked_tool_names = self.rag_infer(message, top_k=rag_num * extra_factor)
|
228 |
+
|
229 |
+
picked_tool_names_no_special = [
|
230 |
+
tool for tool in picked_tool_names
|
231 |
+
if tool not in self.special_tools_name
|
232 |
+
]
|
233 |
picked_tool_names = picked_tool_names_no_special[:rag_num]
|
234 |
|
235 |
picked_tools = self.tooluniverse.get_tool_by_name(picked_tool_names)
|
236 |
picked_tools_prompt = self.tooluniverse.prepare_tool_prompts(picked_tools)
|
237 |
logger.debug("Retrieved %d tools via RAG", len(picked_tools_prompt))
|
238 |
+
|
239 |
if return_call_result:
|
240 |
return picked_tools_prompt, picked_tool_names
|
241 |
return picked_tools_prompt
|
242 |
|
243 |
+
def add_special_tools(self, tools: List, call_agent: bool = False) -> List:
|
244 |
+
"""Add special tools (Finish and optionally CallAgent) to the tools list."""
|
245 |
if self.enable_finish:
|
246 |
tools.append(self.tooluniverse.get_one_tool_by_one_name('Finish', return_prompt=True))
|
247 |
logger.debug("Finish tool added")
|
|
|
250 |
logger.debug("CallAgent tool added")
|
251 |
return tools
|
252 |
|
253 |
+
def set_system_prompt(self, conversation: List[Dict], sys_prompt: str) -> List[Dict]:
|
254 |
+
"""Set or update the system prompt in the conversation."""
|
|
|
|
|
|
|
|
|
255 |
if not conversation:
|
256 |
conversation.append({"role": "system", "content": sys_prompt})
|
257 |
else:
|
258 |
conversation[0] = {"role": "system", "content": sys_prompt}
|
259 |
return conversation
|
260 |
|
261 |
+
def run_function_call(self,
|
262 |
+
fcall_str: str,
|
263 |
+
return_message: bool = False,
|
264 |
+
existing_tools_prompt: Optional[List] = None,
|
265 |
+
message_for_call_agent: Optional[str] = None,
|
266 |
+
call_agent: bool = False,
|
267 |
+
call_agent_level: Optional[int] = None,
|
268 |
+
temperature: Optional[float] = None) -> Tuple[List[Dict], List, str]:
|
269 |
+
"""
|
270 |
+
Execute function calls from the model's output.
|
271 |
+
|
272 |
+
Args:
|
273 |
+
fcall_str: The function call string from the model
|
274 |
+
return_message: Whether to return the message part
|
275 |
+
existing_tools_prompt: Existing tools to consider
|
276 |
+
message_for_call_agent: Original message for CallAgent
|
277 |
+
call_agent: Whether CallAgent is enabled
|
278 |
+
call_agent_level: Current CallAgent level
|
279 |
+
temperature: Temperature for sub-agent calls
|
280 |
+
|
281 |
+
Returns:
|
282 |
+
Tuple of (revised_messages, tools_prompt, special_tool_call)
|
283 |
+
"""
|
284 |
try:
|
285 |
function_call_json, message = self.tooluniverse.extract_function_call_json(
|
286 |
fcall_str, return_message=return_message, verbose=False)
|
|
|
299 |
special_tool_call = 'Finish'
|
300 |
break
|
301 |
elif function_call_json[i]["name"] == 'CallAgent':
|
302 |
+
if call_agent_level is not None and call_agent_level < 2 and call_agent:
|
303 |
solution_plan = function_call_json[i]['arguments']['solution']
|
304 |
full_message = (
|
305 |
+
(message_for_call_agent or "") +
|
306 |
"\nYou must follow the following plan to answer the question: " +
|
307 |
str(solution_plan)
|
|
|
308 |
call_result = self.run_multistep_agent(
|
309 |
full_message, temperature=temperature,
|
310 |
max_new_tokens=512, max_token=131072,
|
|
|
322 |
logger.info("Tool Call Result: %s", call_result)
|
323 |
call_results.append({
|
324 |
"role": "tool",
|
325 |
+
"content": json.dumps({
|
326 |
+
"tool_name": function_call_json[i]["name"],
|
327 |
+
"content": call_result,
|
328 |
+
"call_id": call_id
|
329 |
+
})
|
330 |
})
|
331 |
else:
|
332 |
call_results.append({
|
|
|
339 |
"content": message.strip(),
|
340 |
"tool_calls": json.dumps(function_call_json)
|
341 |
}] + call_results
|
342 |
+
return revised_messages, existing_tools_prompt or [], special_tool_call
|
343 |
+
|
344 |
+
def llm_infer(self,
|
345 |
+
messages: List[Dict],
|
346 |
+
temperature: float = 0.1,
|
347 |
+
tools: Optional[List] = None,
|
348 |
+
output_begin_string: Optional[str] = None,
|
349 |
+
max_new_tokens: int = 512,
|
350 |
+
max_token: int = 131072,
|
351 |
+
skip_special_tokens: bool = True,
|
352 |
+
model: Optional[LLM] = None,
|
353 |
+
check_token_status: bool = False) -> Union[str, Tuple[str, bool]]:
|
354 |
+
"""
|
355 |
+
Perform inference using the LLM.
|
356 |
+
|
357 |
+
Args:
|
358 |
+
messages: Conversation history
|
359 |
+
temperature: Sampling temperature
|
360 |
+
tools: List of tools to include
|
361 |
+
output_begin_string: Prefix for output
|
362 |
+
max_new_tokens: Maximum new tokens to generate
|
363 |
+
max_token: Maximum total tokens allowed
|
364 |
+
skip_special_tokens: Whether to skip special tokens
|
365 |
+
model: Optional custom model to use
|
366 |
+
check_token_status: Whether to check token limits
|
367 |
+
|
368 |
+
Returns:
|
369 |
+
Generated text or tuple with text and overflow flag if check_token_status
|
370 |
+
"""
|
371 |
+
model = model or self.model
|
372 |
+
tokenizer = self.tokenizer
|
373 |
|
374 |
+
sampling_params = SamplingParams(
|
375 |
+
temperature=temperature,
|
376 |
+
max_tokens=max_new_tokens,
|
377 |
+
seed=self.seed,
|
378 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
379 |
|
380 |
+
prompt = self.chat_template.render(
|
381 |
+
messages=messages, tools=tools, add_generation_prompt=True)
|
382 |
+
if output_begin_string is not None:
|
383 |
+
prompt += output_begin_string
|
|
|
|
|
|
|
|
|
384 |
|
385 |
+
token_overflow = False
|
386 |
+
if check_token_status and max_token is not None:
|
387 |
+
num_input_tokens = len(tokenizer.encode(prompt, add_special_tokens=False))
|
388 |
+
logger.info("Input prompt tokens: %d, max_token: %d", num_input_tokens, max_token)
|
389 |
+
if num_input_tokens > max_token:
|
390 |
+
torch.cuda.empty_cache()
|
391 |
+
gc.collect()
|
392 |
+
logger.warning("Token overflow: %d > %d", num_input_tokens, max_token)
|
393 |
+
return (None, True) if check_token_status else None
|
|
|
|
|
|
|
|
|
|
|
|
|
394 |
|
395 |
+
try:
|
396 |
+
output = model.generate(prompt, sampling_params=sampling_params)
|
397 |
+
output_text = output[0].outputs[0].text
|
398 |
+
output_tokens = len(tokenizer.encode(output_text, add_special_tokens=False))
|
399 |
+
logger.debug("Inference output: %s (output tokens: %d)", output_text[:100], output_tokens)
|
400 |
+
|
401 |
+
if skip_special_tokens:
|
402 |
+
output_text = output_text.replace("</s>", "").strip()
|
403 |
+
|
404 |
+
torch.cuda.empty_cache()
|
405 |
+
gc.collect()
|
406 |
+
|
407 |
+
return (output_text, token_overflow) if check_token_status else output_text
|
408 |
+
except Exception as e:
|
409 |
+
logger.error("Inference failed: %s", str(e))
|
410 |
+
raise RuntimeError(f"Inference failed: {str(e)}")
|
411 |
+
|
412 |
+
def run_multistep_agent(self,
|
413 |
+
message: str,
|
414 |
+
temperature: float,
|
415 |
+
max_new_tokens: int,
|
416 |
+
max_token: int,
|
417 |
+
max_round: int = 5,
|
418 |
+
call_agent: bool = False,
|
419 |
+
call_agent_level: int = 0) -> Optional[str]:
|
420 |
+
"""
|
421 |
+
Run multi-step reasoning with the agent.
|
422 |
+
|
423 |
+
Args:
|
424 |
+
message: Input message
|
425 |
+
temperature: Sampling temperature
|
426 |
+
max_new_tokens: Max new tokens per step
|
427 |
+
max_token: Max total tokens
|
428 |
+
max_round: Maximum reasoning rounds
|
429 |
+
call_agent: Whether to enable CallAgent
|
430 |
+
call_agent_level: Current CallAgent level
|
431 |
+
|
432 |
+
Returns:
|
433 |
+
Final answer or None if failed
|
434 |
+
"""
|
435 |
logger.info("Starting multistep agent for message: %s", message[:100])
|
436 |
picked_tools_prompt, call_agent_level = self.initialize_tools_prompt(
|
437 |
call_agent, call_agent_level, message)
|
|
|
501 |
conversation, temperature, max_new_tokens, max_token)
|
502 |
return None
|
503 |
|
504 |
+
def analyze_document(self,
|
505 |
+
file_path: str,
|
506 |
+
temperature: float = 0.1,
|
507 |
+
max_new_tokens: int = 2048,
|
508 |
+
max_token: int = 131072) -> Dict[str, Union[str, List]]:
|
509 |
+
"""
|
510 |
+
Analyze a document and return structured results.
|
511 |
+
|
512 |
+
Args:
|
513 |
+
file_path: Path to the document
|
514 |
+
temperature: Sampling temperature
|
515 |
+
max_new_tokens: Max new tokens per step
|
516 |
+
max_token: Max total tokens
|
517 |
+
|
518 |
+
Returns:
|
519 |
+
Dictionary with analysis results
|
520 |
+
"""
|
521 |
+
logger.info("Starting document analysis for: %s", file_path)
|
522 |
+
start_time = time.time()
|
523 |
+
|
524 |
+
try:
|
525 |
+
extracted_text = self.extract_text(file_path)
|
526 |
+
if not extracted_text:
|
527 |
+
raise ValueError("Could not extract text from document")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
528 |
|
529 |
+
chunks = self.split_text(extracted_text)
|
530 |
+
batches = self.batch_chunks(chunks, batch_size=1)
|
531 |
+
batch_results = []
|
532 |
+
|
533 |
+
for batch in batches:
|
534 |
+
prompt = "\n\n".join(self.build_prompt(chunk) for chunk in batch)
|
535 |
+
response = self.run_multistep_agent(
|
536 |
+
prompt,
|
537 |
+
temperature=temperature,
|
538 |
+
max_new_tokens=max_new_tokens,
|
539 |
+
max_token=max_token,
|
540 |
+
call_agent=False
|
541 |
+
)
|
542 |
+
batch_results.append(self.clean_response(response or "No response"))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
543 |
|
544 |
+
combined = "\n\n".join([res for res in batch_results if not res.startswith("❌")])
|
545 |
+
if not combined:
|
546 |
+
raise ValueError("No valid batch responses generated")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
547 |
|
548 |
+
final_summary = self.generate_final_summary(self, combined)
|
549 |
+
|
550 |
+
return {
|
551 |
+
"status": "success",
|
552 |
+
"summary": final_summary,
|
553 |
+
"batch_results": batch_results,
|
554 |
+
"processing_time": time.time() - start_time
|
555 |
+
}
|
556 |
+
|
557 |
+
except Exception as e:
|
558 |
+
logger.error("Document analysis failed: %s", str(e))
|
559 |
+
return {
|
560 |
+
"status": "error",
|
561 |
+
"message": str(e),
|
562 |
+
"processing_time": time.time() - start_time
|
563 |
+
}
|
564 |
+
|
565 |
+
def get_answer_based_on_unfinished_reasoning(self,
|
566 |
+
conversation: List[Dict],
|
567 |
+
temperature: float,
|
568 |
+
max_new_tokens: int,
|
569 |
+
max_token: int) -> str:
|
570 |
+
"""
|
571 |
+
Generate a final answer when reasoning is incomplete.
|
572 |
+
|
573 |
+
Args:
|
574 |
+
conversation: Current conversation history
|
575 |
+
temperature: Sampling temperature
|
576 |
+
max_new_tokens: Max new tokens
|
577 |
+
max_token: Max total tokens
|
578 |
+
|
579 |
+
Returns:
|
580 |
+
Final answer string
|
581 |
+
"""
|
582 |
+
if conversation[-1]['role'] == 'assistant':
|
583 |
+
conversation.append(
|
584 |
+
{'role': 'tool', 'content': 'Errors occurred during function call; provide final answer with current information.'})
|
585 |
+
finish_tools_prompt = self.add_finish_tools([])
|
586 |
+
last_outputs_str = self.llm_infer(
|
587 |
messages=conversation,
|
588 |
temperature=temperature,
|
589 |
+
tools=finish_tools_prompt,
|
590 |
+
output_begin_string='[FinalAnswer]',
|
591 |
+
skip_special_tokens=True,
|
592 |
max_new_tokens=max_new_tokens,
|
593 |
max_token=max_token)
|
594 |
+
logger.info("Unfinished reasoning answer: %s", last_outputs_str[:100])
|
595 |
+
return last_outputs_str
|
596 |
|
597 |
+
def update_parameters(self, **kwargs) -> Dict:
|
598 |
+
"""
|
599 |
+
Update agent parameters dynamically.
|
600 |
+
|
601 |
+
Args:
|
602 |
+
kwargs: Parameter names and values to update
|
603 |
+
|
604 |
+
Returns:
|
605 |
+
Dictionary of updated parameters
|
606 |
+
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
607 |
updated_attributes = {}
|
608 |
for key, value in kwargs.items():
|
609 |
if hasattr(self, key):
|
|
|
612 |
logger.info("Updated parameters: %s", updated_attributes)
|
613 |
return updated_attributes
|
614 |
|
615 |
+
def cleanup(self) -> None:
|
616 |
+
"""Clean up resources and clear memory."""
|
617 |
+
if hasattr(self, 'model'):
|
618 |
+
del self.model
|
619 |
+
if hasattr(self, 'rag_model'):
|
620 |
+
del self.rag_model
|
621 |
+
if hasattr(self, 'tooluniverse'):
|
622 |
+
del self.tooluniverse
|
623 |
+
torch.cuda.empty_cache()
|
624 |
+
gc.collect()
|
625 |
+
logger.info("TxAgent resources cleaned up")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
626 |
|
627 |
+
def __del__(self):
|
628 |
+
"""Destructor to ensure proper cleanup."""
|
629 |
+
self.cleanup()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|