File size: 16,690 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 |
import json
import os
import traceback
from abc import ABC, abstractmethod
from copy import deepcopy
from dataclasses import dataclass, field
from datetime import datetime
import jsonlines
import PIL.Image
import torch
from FlagEmbedding import BGEM3FlagModel
from jinja2 import Environment, StrictUndefined
from rich import print
from src.apis import API_TYPES, CodeExecutor
from src.llms import Role
from src.model_utils import get_text_embedding
from src.presentation import Presentation, SlidePage
from src.utils import Config, get_slide_content, pexists, pjoin, tenacity
@dataclass
class PPTGen(ABC):
"""
Stage II: Presentation Generation
An abstract base class for generating PowerPoint presentations.
It accepts a reference presentation as input, then generates a presentation outline and slides.
"""
roles: list[str] = field(default_factory=list)
def __init__(
self,
text_model: BGEM3FlagModel,
retry_times: int = 3,
force_pages: bool = False,
error_exit: bool = True,
record_cost: bool = True,
**kwargs,
):
"""
Initialize the PPTGen.
Args:
text_model (BGEM3FlagModel): The text model for generating content.
retry_times (int): The number of times to retry failed actions.
force_pages (bool): Whether to force a specific number of pages.
error_exit (bool): Whether to exit on error.
record_cost (bool): Whether to record the cost of generation.
**kwargs: Additional arguments.
"""
self.text_model = text_model
self.retry_times = retry_times
self.force_pages = force_pages
self.error_exit = error_exit
self._hire_staffs(record_cost, **kwargs)
def set_reference(
self,
presentation: Presentation,
slide_induction: dict,
):
"""
Set the reference presentation and extracted presentation information.
Args:
presentation (Presentation): The presentation object.
slide_induction (dict): The slide induction data.
Returns:
PPTGen: The updated PPTGen object.
"""
self.presentation = presentation
self.slide_induction = slide_induction
self.functional_keys = slide_induction.pop("functional_keys")
self.layout_names = list(slide_induction.keys())
self.layout_embeddings = torch.stack(
get_text_embedding(self.layout_names, self.text_model)
)
self.empty_prs = deepcopy(presentation)
return self
def generate_pres(
self,
config: Config,
images: dict[str, str],
num_slides: int,
doc_json: dict[str, str],
):
"""
Generate a PowerPoint presentation.
Args:
config (Config): The configuration object.
images (dict[str, str]): A dictionary of image paths and captions.
num_slides (int): The number of slides to generate.
doc_json (dict[str, str]): The document JSON data.
Save:
final.pptx: The final PowerPoint presentation to the config.RUN_DIR directory.
Raise:
ValueError: if failed to generate presentation outline.
"""
self.config = config
self.doc_json = doc_json
meta_data = "\n".join(
[f"{k}: {v}" for k, v in self.doc_json.get("metadata", {}).items()]
)
self.metadata = (
f"{meta_data}\nPresentation Time: {datetime.now().strftime('%Y-%m-%d')}\n"
)
self.image_information = ""
for k, v in images.items():
assert pexists(k), f"Image {k} not found"
size = PIL.Image.open(k).size
self.image_information += (
f"Image path: {k}, size: {size[0]}*{size[1]} px\n caption: {v}\n"
)
succ_flag = True
code_executor = CodeExecutor(self.retry_times)
self.outline = self._generate_outline(num_slides)
self.simple_outline = "\n".join(
[
f"Slide {slide_idx+1}: {slide_title}"
for slide_idx, slide_title in enumerate(self.outline)
]
)
generated_slides = []
for slide_data in enumerate(self.outline.items()):
if self.force_pages and slide_data[0] == num_slides:
break
slide = self._generate_slide(slide_data, code_executor)
if slide is not None:
generated_slides.append(slide)
continue
if self.error_exit:
succ_flag = False
break
self._save_history(code_executor)
if succ_flag:
self.empty_prs.slides = generated_slides
self.empty_prs.save(pjoin(self.config.RUN_DIR, "final.pptx"))
def _save_history(self, code_executor: CodeExecutor):
"""
Save the history of code execution, API calls and agent steps.
"""
os.makedirs(pjoin(self.config.RUN_DIR, "history"), exist_ok=True)
for role in self.staffs.values():
role.save_history(pjoin(self.config.RUN_DIR, "history"))
role.history = []
if len(code_executor.code_history) == 0:
return
with jsonlines.open(
pjoin(self.config.RUN_DIR, "code_steps.jsonl"), "w"
) as writer:
writer.write_all(code_executor.code_history)
with jsonlines.open(
pjoin(self.config.RUN_DIR, "agent_steps.jsonl"), "w"
) as writer:
writer.write_all(code_executor.api_history)
@tenacity
def _generate_outline(self, num_slides: int):
"""
Generate an outline for the presentation.
Args:
num_slides (int): The number of slides to generate.
Returns:
dict: The generated outline.
"""
outline_file = pjoin(self.config.RUN_DIR, "presentation_outline.json")
doc_overview = deepcopy(self.doc_json)
for section in doc_overview["sections"]:
[sub.pop("content") for sub in section["subsections"]]
if pexists(outline_file):
outline = json.load(open(outline_file, "r"))
else:
outline = self.staffs["planner"](
num_slides=num_slides,
layouts="\n".join(
set(self.slide_induction.keys()).difference(self.functional_keys)
),
functional_keys="\n".join(self.functional_keys),
json_content=doc_overview,
image_information=self.image_information,
)
outline = self._valid_outline(outline)
json.dump(
outline,
open(outline_file, "w"),
ensure_ascii=False,
indent=4,
)
return outline
def _valid_outline(self, outline: dict, retry: int = 0) -> dict:
"""
Validate the generated outline.
Raises:
ValueError: If the outline is invalid.
"""
try:
for slide in outline.values():
layout_sim = torch.cosine_similarity(
get_text_embedding(slide["layout"], self.text_model),
self.layout_embeddings,
)
if layout_sim.max() < 0.7:
raise ValueError(
f"Layout `{slide['layout']}` not found, must be one of {self.layout_names}"
)
slide["layout"] = self.layout_names[layout_sim.argmax().item()]
if any(
not {"layout", "subsections", "description"}.issubset(set(slide.keys()))
for slide in outline.values()
):
raise ValueError(
"Invalid outline structure, must be a dict with layout, subsections, description"
)
except ValueError as e:
print(outline, e)
if retry < self.retry_times:
new_outline = self.staffs["planner"].retry(
str(e), traceback.format_exc(), retry + 1
)
return self._valid_outline(new_outline, retry + 1)
else:
raise ValueError("Failed to generate outline, tried too many times")
return outline
def _hire_staffs(self, record_cost: bool, **kwargs) -> dict[str, Role]:
"""
Initialize agent roles and their models
"""
jinja_env = Environment(undefined=StrictUndefined)
self.staffs = {
role: Role(
role,
env=jinja_env,
record_cost=record_cost,
text_model=self.text_model,
**kwargs,
)
for role in ["planner"] + self.roles
}
@abstractmethod
def synergize(
self,
template: dict,
slide_content: str,
code_executor: CodeExecutor,
image_info: str,
) -> SlidePage:
"""
Synergize Agents to generate a slide.
Returns:
SlidePage: The generated slide.
"""
pass
def _generate_slide(self, slide_data, code_executor: CodeExecutor) -> SlidePage:
"""
Generate a slide from the slide data.
"""
slide_idx, (slide_title, slide) = slide_data
images_info = "No Images"
if any(
[
i in slide["layout"]
for i in ["picture", "chart", "table", "diagram", "freeform"]
]
):
images_info = self.image_information
slide_content = f"Slide-{slide_idx+1} " + get_slide_content(
self.doc_json, slide_title, slide
)
template = deepcopy(self.slide_induction[slide["layout"]])
try:
return self.synergize(
template,
slide_content,
code_executor,
images_info,
)
except Exception as e:
print(f"generate slide {slide_idx} failed: {e}")
print(traceback.format_exc())
print(self.config.RUN_DIR)
# 价格scale factor
class PPTCrew(PPTGen):
"""
A class to generate PowerPoint presentations with a crew of agents.
"""
roles: list[str] = ["editor", "coder"]
def synergize(
self,
template: dict,
slide_content: str,
code_executor: CodeExecutor,
images_info: str,
) -> SlidePage:
"""
Synergize Agents to generate a slide.
Args:
template (dict): The template data.
slide_content (str): The slide content.
code_executor (CodeExecutor): The code executor object.
images_info (str): The image information.
Returns:
SlidePage: The generated slide.
"""
content_schema = template["content_schema"]
old_data = self._prepare_schema(content_schema)
editor_output = self.staffs["editor"](
schema=content_schema,
outline=self.simple_outline,
metadata=self.metadata,
text=slide_content,
images_info=images_info,
)
command_list = self._generate_commands(editor_output, content_schema, old_data)
edit_actions = self.staffs["coder"](
api_docs=code_executor.get_apis_docs(API_TYPES.Agent.value),
edit_target=self.presentation.slides[template["template_id"] - 1].to_html(),
command_list="\n".join([str(i) for i in command_list]),
)
for error_idx in range(self.retry_times):
edited_slide: SlidePage = deepcopy(
self.presentation.slides[template["template_id"] - 1]
)
feedback = code_executor.execute_actions(edit_actions, edited_slide)
if feedback is None:
break
if error_idx == self.retry_times:
raise Exception(
f"Failed to generate slide, tried too many times at editing\ntraceback: {feedback[1]}"
)
edit_actions = self.staffs["coder"].retry(*feedback, error_idx + 1)
self.empty_prs.build_slide(edited_slide)
return edited_slide
def _prepare_schema(self, content_schema: dict):
"""
Prepare the content schema for editing.
Args:
content_schema (dict): The content schema.
Returns:
dict: The old data extracted from the schema.
"""
old_data = {}
for el_name, el_info in content_schema.items():
if el_info["type"] == "text":
if not isinstance(el_info["data"], list):
el_info["data"] = [el_info["data"]]
if len(el_info["data"]) > 1:
charater_counts = [len(i) for i in el_info["data"]]
content_schema[el_name]["suggestedCharacters"] = (
str(min(charater_counts)) + "-" + str(max(charater_counts))
)
else:
content_schema[el_name]["suggestedCharacters"] = "<" + str(
len(el_info["data"][0])
)
old_data[el_name] = el_info.pop("data")
content_schema[el_name]["default_quantity"] = 1
if isinstance(old_data[el_name], list):
content_schema[el_name]["default_quantity"] = len(old_data[el_name])
assert len(old_data) > 0, "No old data generated"
return old_data
def _generate_commands(
self, editor_output: dict, content_schema: dict, old_data: dict, retry: int = 0
):
"""
Generate commands for editing the slide content.
Args:
editor_output (dict): The editor output.
content_schema (dict): The content schema.
old_data (dict): The old data.
retry (int): The number of retries.
Returns:
list: A list of commands.
Raises:
Exception: If command generation fails.
"""
command_list = []
try:
for el_name, el_data in editor_output.items():
assert (
"data" in el_data
), """key `data` not found in output
please give your output as a dict like
{
"element1": {
"data": ["text1", "text2"] for text elements
or ["/path/to/image", "..."] for image elements
},
}"""
charater_counts = [len(i) for i in el_data["data"]]
max_charater_count = max([len(i) for i in old_data[el_name]])
if max(charater_counts) > max_charater_count * 1.5:
raise ValueError(
f"Content for '{el_name}' exceeds character limit ({max(charater_counts)} > {max_charater_count}). "
f"Please reduce the content length to maintain slide readability and visual balance. "
f"Current text: '{el_data['data']}'"
)
except Exception as e:
if retry < self.retry_times:
new_output = self.staffs["editor"].retry(
e,
traceback.format_exc(),
retry + 1,
)
return self._generate_commands(
new_output, content_schema, old_data, retry + 1
)
for el_name, old_content in old_data.items():
if not isinstance(old_content, list):
old_content = [old_content]
new_content = editor_output.get(el_name, {}).get("data", None)
if not isinstance(new_content, list):
new_content = [new_content]
new_content = [i for i in new_content if i]
if content_schema[el_name]["type"] == "image":
new_content = [i for i in new_content if pexists(i)]
quantity_change = len(new_content) - len(old_content)
command_list.append(
(
el_name,
content_schema[el_name]["type"],
f"quantity_change: {quantity_change}",
old_content,
new_content,
)
)
assert len(command_list) > 0, "No commands generated"
return command_list
|