File size: 21,316 Bytes
fcaa164 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 |
"""
Content planner for project page generation.
Plans the structure and content organization for the project page.
"""
import json
import yaml
import os
from jinja2 import Environment, StrictUndefined
from camel.models import ModelFactory
from camel.agents import ChatAgent
from utils.wei_utils import account_token
from utils.src.utils import get_json_from_response
from camel.messages import BaseMessage
from rich import print
from rich.pretty import Pretty
import base64
from camel.messages import BaseMessage
from camel.models import ModelFactory
def filter_references(md_content: str) -> str:
lines = md_content.splitlines()
result_lines = []
for line in lines:
if line.strip().lower().startswith("## references"):
break
result_lines.append(line)
return "\n".join(result_lines)
class ProjectPageContentPlanner:
"""Plans the content structure and organization for project pages."""
def __init__(self, agent_config, args):
self.agent_config = agent_config
self.args = args
self.planner_agent = self._create_planner_agent()
self.reviewer_agent = self._create_reviewer_agent()
os.makedirs('project_contents', exist_ok=True)
def _create_planner_agent(self):
"""Create the content planning (generation) agent."""
model_type = str(self.agent_config['model_type'])
# Get API key from environment variables
api_key = None
if self.args.model_name_t in ['4o', '4o-mini', 'gpt-4.1', 'gpt-4.1-mini', 'o1', 'o3', 'o3-mini']:
api_key = os.environ.get('OPENAI_API_KEY')
elif self.args.model_name_t in ['gemini', 'gemini-2.5-pro', 'gemini-2.5-flash']:
api_key = os.environ.get('GEMINI_API_KEY')
elif self.args.model_name_t in ['qwen', 'qwen-plus', 'qwen-max', 'qwen-long']:
api_key = os.environ.get('QWEN_API_KEY')
elif self.args.model_name_t.startswith('openrouter_'):
api_key = os.environ.get('OPENROUTER_API_KEY')
elif self.args.model_name_t in ['zhipuai']:
api_key = os.environ.get('ZHIPUAI_API_KEY')
if model_type.startswith('vllm_qwen') or 'vllm' in model_type.lower():
model = ModelFactory.create(
model_platform=self.agent_config['model_platform'],
model_type=self.agent_config['model_type'],
model_config_dict=self.agent_config['model_config'],
url=self.agent_config.get('url', None),
api_key=api_key,
)
else:
model = ModelFactory.create(
model_platform=self.agent_config['model_platform'],
model_type=self.agent_config['model_type'],
model_config_dict=self.agent_config['model_config'],
api_key=api_key,
)
system_message = """You are a helpful academic expert and web developer, who is specialized in generating a paper project page, from given research paper's contents and figures."""
return ChatAgent(
system_message=system_message,
model=model,
message_window_size=10,
token_limit=self.agent_config.get('token_limit', None)
)
def _create_reviewer_agent(self):
model_type = str(self.agent_config['model_type'])
# Get API key from environment variables
api_key = None
if self.args.model_name_t in ['4o', '4o-mini', 'gpt-4.1', 'gpt-4.1-mini', 'o1', 'o3', 'o3-mini']:
api_key = os.environ.get('OPENAI_API_KEY')
elif self.args.model_name_t in ['gemini', 'gemini-2.5-pro', 'gemini-2.5-flash']:
api_key = os.environ.get('GEMINI_API_KEY')
elif self.args.model_name_t in ['qwen', 'qwen-plus', 'qwen-max', 'qwen-long']:
api_key = os.environ.get('QWEN_API_KEY')
elif self.args.model_name_t.startswith('openrouter_'):
api_key = os.environ.get('OPENROUTER_API_KEY')
elif self.args.model_name_t in ['zhipuai']:
api_key = os.environ.get('ZHIPUAI_API_KEY')
if model_type.startswith('vllm_qwen') or 'vllm' in model_type.lower():
model = ModelFactory.create(
model_platform=self.agent_config['model_platform'],
model_type=self.agent_config['model_type'],
model_config_dict=self.agent_config['model_config'],
url=self.agent_config.get('url', None),
api_key=api_key,
)
else:
model = ModelFactory.create(
model_platform=self.agent_config['model_platform'],
model_type=self.agent_config['model_type'],
model_config_dict=self.agent_config['model_config'],
api_key=api_key,
)
reviewer_system = (
"You are a precise, constructive reviewer of generated project pages. "
)
return ChatAgent(
system_message=reviewer_system,
model=model,
message_window_size=10,
token_limit=self.agent_config.get('token_limit', None)
)
def _render_generation_prompt(self, paper_content, figures, text_page_content, template_str):
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(template_str)
jinja_args = {
'paper_content': paper_content,
'figures': json.dumps(figures, indent=2),
'project_page_content': json.dumps(text_page_content, indent=2),
}
return template.render(**jinja_args)
def _build_reviewer_prompt(self, paper_content, figures, text_page_content, generated_json):
with open('utils/prompt_templates/page_templates/full_content_review.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
jinja_args = {
'paper_content': paper_content,
'figures': json.dumps(figures['images'], indent=2),
'tables': json.dumps(figures['tables'], indent=2),
"generated_content": generated_json
}
prompt = template.render(**jinja_args)
return prompt
def _build_revision_prompt(self, review_json):
with open('utils/prompt_templates/page_templates/full_content_revise.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
jinja_args = {
"review_content": json.dumps(review_json, indent=2)
}
prompt = template.render(**jinja_args)
return prompt
def _build_revision_prompt_with_resume(self, review_json, current_content, figures):
with open('utils/prompt_templates/page_templates/full_content_revise_with_resume.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
print(review_json)
jinja_args = {
"review_content": json.dumps(review_json, indent=2),
"figures": json.dumps(figures, indent=2),
"current_content": current_content
}
prompt = template.render(**jinja_args)
return prompt
def full_content_generation(
self,
args,
paper_content,
figures,
generated_section,
text_page_content,
):
"""
Plan + Generate -> Review -> Revise
Args:
paper_content: parsed paper content
figures: list/dict of figures
generated_section: format_instructions / schema hints
text_page_content: initial text-only page structure
Returns:
tuple: (final_generated_content_json, input_token_total, output_token_total)
"""
if args.resume in ['parse_pdf','generate_content']:
print("full content generation start")
with open('utils/prompt_templates/page_templates/full_content_generation.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
jinja_args = {
'paper_content': paper_content,
'figures': json.dumps(figures, indent=2),
'project_page_content': json.dumps(text_page_content, indent=2)
}
prompt = template.render(**jinja_args)
self.planner_agent.reset()
response = self.planner_agent.step(prompt)
gen_in_tok, gen_out_tok = account_token(response)
current_output = get_json_from_response(response.msgs[0].content)
first_path = f'project_contents/{self.args.paper_name}_generated_full_content.v0.json'
with open(first_path, 'w', encoding='utf-8') as f:
json.dump(current_output, f, ensure_ascii=False, indent=2)
print(f" - Initial generation saved: {first_path}")
total_in_tok, total_out_tok = gen_in_tok, gen_out_tok
else:
print("Skipping initial full content generation, loading existing content.")
with open(f'project_contents/{self.args.paper_name}_generated_full_content.v0.json', 'r', encoding='utf-8') as f:
current_output = json.load(f)
total_in_tok, total_out_tok = 0, 0
for it in range(0, args.full_content_check_times):
# check
self.reviewer_agent.reset()
review_prompt = self._build_reviewer_prompt(
paper_content=paper_content,
figures=figures,
text_page_content=text_page_content,
generated_json=current_output
)
review_resp = self.reviewer_agent.step(review_prompt)
rin, rout = account_token(review_resp)
review_json = get_json_from_response(review_resp.msgs[0].content)
review_path = f'project_contents/{self.args.paper_name}_review.iter{it}.json'
with open(review_path, 'w', encoding='utf-8') as f:
json.dump(review_json, f, ensure_ascii=False, indent=2)
print(f" - Review saved: {review_path}")
total_in_tok += rin
total_out_tok += rout
if args.resume != 'full_content_check':
revision_prompt = self._build_revision_prompt(
review_json=review_json
)
else:
revision_prompt = self._build_revision_prompt_with_resume(
review_json=review_json,
current_content=current_output,
figures=figures
)
rev_resp = self.planner_agent.step(revision_prompt)
rin2, rout2 = account_token(rev_resp)
revised_output = get_json_from_response(rev_resp.msgs[0].content)
out_path = f'project_contents/{self.args.paper_name}_generated_full_content.v{it+1}.json'
with open(out_path, 'w', encoding='utf-8') as f:
json.dump(revised_output, f, ensure_ascii=False, indent=2)
print(f" - Revised generation saved: {out_path}")
total_in_tok += rin2
total_out_tok += rout2
current_output = revised_output
if self.args.human_input == '1':
print('-'*50)
print(Pretty(current_output, expand_all=True))
print('-'*50)
user_feedback = input('The above is the final generated full content! If you are satisfied with the generated content, enter yes\n If not, enter your feedback.\n')
while user_feedback.lower() != 'yes':
message = BaseMessage.make_assistant_message(
role_name='User',
content='human feedback'+user_feedback +"The above is human feedback. Please make modifications based on this feedback and the original content.The output format is as specified above."
)
response = self.planner_agent.step(message)
current_output = get_json_from_response(response.msgs[0].content)
print('-'*50)
print(Pretty(current_output, expand_all=True))
print('-'*50)
user_feedback = input('The above is the final generated full content! If you are satisfied with the generated content, enter yes. \n If not, enter your feedback.\n')
in_tok, out_tok = account_token(response)
total_in_tok += in_tok
total_out_tok += out_tok
# 4) 最终保存(保持你原有的命名)
final_path = f'project_contents/{self.args.paper_name}_generated_full_content.json'
with open(final_path, 'w', encoding='utf-8') as f:
json.dump(current_output, f, ensure_ascii=False, indent=2)
print(f"full content generation completed. Tokens: {total_in_tok} -> {total_out_tok}")
print(f" - Final content: {final_path}")
return current_output, total_in_tok, total_out_tok
def section_generation(self, paper_content, figures):
"""
Plan the content structure for the project page.
Args:
paper_content: Parsed paper content
Returns:
dict: project page content
"""
# Load planning prompt template
with open('utils/prompt_templates/page_templates/section_generation.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
json_format_example = """
```json
{{
"Introduction": "Brief overview of the paper's main topic and objectives.",
"Methodology": "Description of the methods used in the research.",
"Results": "Summary of the key findings and results."
}}
```
"""
# Prepare template arguments
jinja_args = {
'paper_content': paper_content,
'json_format_example': json.dumps(paper_content, indent=2)
}
prompt = template.render(**jinja_args)
# Generate content plan
self.planner_agent.reset()
response = self.planner_agent.step(prompt)
input_token, output_token = account_token(response)
generated_section = get_json_from_response(response.msgs[0].content)
if self.args.human_input == '1':
print('-'*50)
print(Pretty(generated_section, expand_all=True))
print('-'*50)
user_feedback = input('The above is the generated section! If you are satisfied with the generated section, enter yes. \nIf not, enter your feedback.\n')
while user_feedback.lower() != 'yes':
message = BaseMessage.make_assistant_message(
role_name='User',
content='human feedback'+user_feedback +"The above is human feedback. Please make modifications based on this feedback and the original content.The output format is as specified above."
)
response = self.planner_agent.step(message)
generated_section = get_json_from_response(response.msgs[0].content)
print('-'*50)
print(Pretty(generated_section, expand_all=True))
print('-'*50)
user_feedback = input('The above is the generated section! If you are satisfied with the generated section, enter yes. \nIf not, enter your feedback.\n')
in_tok, out_tok = account_token(response)
input_token += in_tok
output_token += out_tok
print(f"section planning completed. Tokens: {input_token} -> {output_token}")
def create_dynamic_page_dict(sections: dict[str, str]) -> dict[str, str]:
poster_dict = {
"title": "Title of the paper",
"authors": "Authors of the paper, Each author must be accompanied by the superscript number(s) of their corresponding affiliation(s).",
"affiliation": "Affiliation of the authors, each affiliation must be accompanied by the corresponding superscript number.",
}
poster_dict.update(sections)
return poster_dict
generated_section = create_dynamic_page_dict(generated_section)
# Save generated content
# print(self.agent_config)
generated_path = f'project_contents/{self.args.paper_name}_generated_section.json'
with open(generated_path, 'w') as f:
json.dump(generated_section, f, indent=4)
print(f" - Generated section plan: {generated_path}")
return generated_section, input_token, output_token
def text_content_generation(self, paper_content, figures, generated_section):
"""
Plan the content structure for the project page.
Args:
paper_content: Parsed paper content
Returns:
dict: project page content
"""
# Delete tags in figures
figures_ = {}
figures_['images'] = [{k: v for k, v in value.items() if k != 'tag'} for value in figures['images'].values()]
figures_['tables'] = [{k: v for k, v in value.items() if k != 'tag'} for value in figures['tables'].values()]
# Load planning prompt template
with open('utils/prompt_templates/page_templates/text_content_generation.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
# Prepare template arguments
jinja_args = {
'paper_content': paper_content,
'figures': json.dumps(figures_, indent=2),
'format_instructions': json.dumps(generated_section, indent=2)
}
prompt = template.render(**jinja_args)
# Generate content plan
self.planner_agent.reset()
response = self.planner_agent.step(prompt)
input_token, output_token = account_token(response)
generated_text_content = get_json_from_response(response.msgs[0].content)
print(f"text content generation completed. Tokens: {input_token} -> {output_token}")
# Save generated content
generated_path = f'project_contents/{self.args.paper_name}_generated_text_content.json'
with open(generated_path, 'w') as f:
json.dump(generated_text_content, f, indent=4)
print(f" - Generated text content: {generated_path}")
return generated_text_content, input_token, output_token
def filter_raw_content(self, paper_content, figures):
paper_content = filter_references(paper_content)
# Load planning prompt template
with open('utils/prompt_templates/page_templates/filter_figures.yaml', 'r') as f:
planner_config = yaml.safe_load(f)
jinja_env = Environment(undefined=StrictUndefined)
template = jinja_env.from_string(planner_config["template"])
# Prepare template arguments
jinja_args = {
'paper_content': paper_content,
'figures': json.dumps(figures, indent=2),
}
prompt = template.render(**jinja_args)
# Generate filtered figures
self.planner_agent.reset()
response = self.planner_agent.step(prompt)
input_token, output_token = account_token(response)
filtered_figures = get_json_from_response(response.msgs[0].content)
#print(filtered_figures)
def remove_items_without_section(data: dict) -> dict:
for key in ["images", "tables"]:
if key in data and isinstance(data[key], dict):
data[key] = {
k: v for k, v in data[key].items()
if v.get("original_section") is not None
}
return data
filtered_figures = remove_items_without_section(filtered_figures)
print(f"filtered figures generation completed. Tokens: {input_token} -> {output_token}")
# Save generated filtered figures
generated_path = f'project_contents/{self.args.paper_name}_generated_filtered_figures.json'
with open(generated_path, 'w') as f:
json.dump(filtered_figures, f, indent=4)
print(f" - Generated filtered figures: {generated_path}")
return paper_content, filtered_figures, input_token, output_token
|