File size: 41,854 Bytes
9b5ca29 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 |
import os
import re
import json
import logging
import glob
from pathlib import Path
from typing import Union, List, Dict, Optional, Tuple, Any
from PIL import Image
from src.utils.utils import extract_json
from mllm_tools.utils import _prepare_text_inputs, _extract_code, _prepare_text_image_inputs
from mllm_tools.gemini import GeminiWrapper
from mllm_tools.vertex_ai import VertexAIWrapper
from task_generator import (
get_prompt_code_generation,
get_prompt_fix_error,
get_prompt_visual_fix_error,
get_banned_reasonings,
get_prompt_rag_query_generation_fix_error,
get_prompt_context_learning_code,
get_prompt_rag_query_generation_code
)
from task_generator.prompts_raw import (
_code_font_size,
_code_disable,
_code_limit,
_prompt_manim_cheatsheet
)
from src.rag.vector_store import RAGVectorStore
# Configuration constants
DEFAULT_MAX_RETRIES = 10
DEFAULT_RAG_K_VALUE = 2
CACHE_FILE_ENCODING = 'utf-8'
CODE_PATTERN = r"```python(.*)```"
JSON_PATTERN = r'```json(.*)```'
# Set up logging
logger = logging.getLogger(__name__)
class CodeGenerator:
"""A class for generating and managing Manim code with improved error handling and maintainability."""
def __init__(
self,
scene_model: Any,
helper_model: Any,
output_dir: str = "output",
print_response: bool = False,
use_rag: bool = False,
use_context_learning: bool = False,
context_learning_path: str = "data/context_learning",
chroma_db_path: str = "rag/chroma_db",
manim_docs_path: str = "rag/manim_docs",
embedding_model: str = "azure/text-embedding-3-large",
use_visual_fix_code: bool = False,
use_langfuse: bool = True,
session_id: Optional[str] = None
) -> None:
"""Initialize the CodeGenerator.
Args:
scene_model: The model used for scene generation
helper_model: The model used for helper tasks
output_dir (str, optional): Directory for output files. Defaults to "output".
print_response (bool, optional): Whether to print model responses. Defaults to False.
use_rag (bool, optional): Whether to use RAG. Defaults to False.
use_context_learning (bool, optional): Whether to use context learning. Defaults to False.
context_learning_path (str, optional): Path to context learning examples. Defaults to "data/context_learning".
chroma_db_path (str, optional): Path to ChromaDB. Defaults to "rag/chroma_db".
manim_docs_path (str, optional): Path to Manim docs. Defaults to "rag/manim_docs".
embedding_model (str, optional): Name of embedding model. Defaults to "azure/text-embedding-3-large".
use_visual_fix_code (bool, optional): Whether to use visual code fixing. Defaults to False.
use_langfuse (bool, optional): Whether to use Langfuse logging. Defaults to True.
session_id (str, optional): Session identifier. Defaults to None.
"""
self.scene_model = scene_model
self.helper_model = helper_model
self.output_dir = Path(output_dir)
self.print_response = print_response
self.use_rag = use_rag
self.use_context_learning = use_context_learning
self.context_learning_path = Path(context_learning_path)
self.manim_docs_path = Path(manim_docs_path)
self.use_visual_fix_code = use_visual_fix_code
self.session_id = session_id
# Ensure output directory exists
self.output_dir.mkdir(parents=True, exist_ok=True)
# Load context examples and banned reasonings
self.context_examples = self._load_context_examples() if use_context_learning else None
self.banned_reasonings = self._load_banned_reasonings()
# Initialize RAG vector store if enabled
self.vector_store = self._initialize_vector_store(
chroma_db_path, embedding_model, use_langfuse
) if use_rag else None
logger.info(f"CodeGenerator initialized with RAG: {use_rag}, Context Learning: {use_context_learning}")
def _load_banned_reasonings(self) -> List[str]:
"""Load banned reasonings with error handling."""
try:
return get_banned_reasonings()
except Exception as e:
logger.warning(f"Failed to load banned reasonings: {e}")
return []
def _initialize_vector_store(self, chroma_db_path: str, embedding_model: str, use_langfuse: bool) -> Optional[RAGVectorStore]:
"""Initialize RAG vector store with error handling."""
try:
return RAGVectorStore(
chroma_db_path=chroma_db_path,
manim_docs_path=str(self.manim_docs_path),
embedding_model=embedding_model,
session_id=self.session_id,
use_langfuse=use_langfuse
)
except Exception as e:
logger.error(f"Failed to initialize RAG vector store: {e}")
return None
def _load_context_examples(self) -> Optional[str]:
"""Load all context learning examples from the specified directory.
Returns:
Optional[str]: Formatted context learning examples, or None if no examples found.
"""
if not self.context_learning_path.exists():
logger.warning(f"Context learning path does not exist: {self.context_learning_path}")
return None
examples = []
pattern = str(self.context_learning_path / "**" / "*.py")
try:
for example_file in glob.glob(pattern, recursive=True):
example_path = Path(example_file)
try:
with example_path.open('r', encoding=CACHE_FILE_ENCODING) as f:
content = f.read()
examples.append(f"# Example from {example_path.name}\n{content}\n")
except (IOError, UnicodeDecodeError) as e:
logger.warning(f"Failed to read example file {example_file}: {e}")
continue
if examples:
formatted_examples = get_prompt_context_learning_code(
examples="\n".join(examples)
)
logger.info(f"Loaded {len(examples)} context learning examples")
return formatted_examples
except Exception as e:
logger.error(f"Error loading context examples: {e}")
return None
def _create_cache_directory(self, topic: str, scene_number: int, cache_type: str = "rag_cache") -> Path:
"""Create and return cache directory path."""
sanitized_topic = re.sub(r'[^a-z0-9_]+', '_', topic.lower())
cache_dir = self.output_dir / sanitized_topic / f"scene{scene_number}" / cache_type
cache_dir.mkdir(parents=True, exist_ok=True)
return cache_dir
def _load_cached_queries(self, cache_file: Path) -> Optional[List[str]]:
"""Load cached queries from file with error handling."""
if not cache_file.exists():
return None
try:
with cache_file.open('r', encoding=CACHE_FILE_ENCODING) as f:
cached_queries = json.load(f)
logger.debug(f"Loaded cached queries from {cache_file}")
return cached_queries
except (json.JSONDecodeError, IOError) as e:
logger.warning(f"Failed to load cached queries from {cache_file}: {e}")
return None
def _save_queries_to_cache(self, queries: List[str], cache_file: Path) -> None:
"""Save queries to cache file with error handling."""
try:
with cache_file.open('w', encoding=CACHE_FILE_ENCODING) as f:
json.dump(queries, f, indent=2)
logger.debug(f"Saved queries to cache: {cache_file}")
except (IOError, TypeError) as e:
logger.error(f"Failed to save queries to cache {cache_file}: {e}")
def _extract_json_from_response(self, response: str, error_context: str = "") -> List[str]:
"""Extract and parse JSON from model response with improved error handling."""
# Try to extract JSON from code blocks first
json_match = re.search(JSON_PATTERN, response, re.DOTALL)
if json_match:
json_text = json_match.group(1).strip()
else:
# Fallback: clean the response and try direct parsing
json_text = response.replace("```json", "").replace("```", "").strip()
try:
return json.loads(json_text)
except json.JSONDecodeError as e:
logger.error(f"JSONDecodeError when parsing {error_context}: {e}")
logger.error(f"Response text was: {response[:500]}...")
return []
def _generate_rag_queries_code(
self,
implementation: str,
scene_trace_id: Optional[str] = None,
topic: Optional[str] = None,
scene_number: Optional[int] = None,
session_id: Optional[str] = None,
relevant_plugins: List[str] = None
) -> List[str]:
"""Generate RAG queries from the implementation plan.
Args:
implementation: The implementation plan text
scene_trace_id: Trace ID for the scene
topic: Topic of the scene
scene_number: Scene number
session_id: Session identifier
relevant_plugins: List of relevant plugins
Returns:
List of generated RAG queries
"""
if relevant_plugins is None:
relevant_plugins = []
if not topic or scene_number is None:
logger.warning("Missing topic or scene_number for RAG query generation")
return []
# Setup cache
cache_dir = self._create_cache_directory(topic, scene_number)
cache_file = cache_dir / "rag_queries_code.json"
# Try to load from cache
cached_queries = self._load_cached_queries(cache_file)
if cached_queries is not None:
logger.info(f"Using cached RAG queries for {topic}_scene{scene_number}")
return cached_queries
# Generate new queries
try:
plugins_text = ", ".join(relevant_plugins) if relevant_plugins else "No plugins are relevant."
prompt = get_prompt_rag_query_generation_code(implementation, plugins_text)
response = self.helper_model(
_prepare_text_inputs(prompt),
metadata={
"generation_name": "rag_query_generation",
"trace_id": scene_trace_id,
"tags": [topic, f"scene{scene_number}"],
"session_id": session_id
}
)
logger.debug(f"RAG queries response: {response[:200]}...")
queries = self._extract_json_from_response(response, "RAG queries for code generation")
# Cache the queries
if queries:
self._save_queries_to_cache(queries, cache_file)
return queries
except Exception as e:
logger.error(f"Error generating RAG queries for code: {e}")
return []
def _generate_rag_queries_error_fix(
self,
error: str,
code: str,
scene_trace_id: Optional[str] = None,
topic: Optional[str] = None,
scene_number: Optional[int] = None,
session_id: Optional[str] = None,
relevant_plugins: List[str] = None
) -> List[str]:
"""Generate RAG queries for fixing code errors.
Args:
error: The error message to fix
code: The code containing the error
scene_trace_id: Trace ID for the scene
topic: Topic of the scene
scene_number: Scene number
session_id: Session identifier
relevant_plugins: List of relevant plugins
Returns:
List of generated RAG queries for error fixing
"""
if relevant_plugins is None:
relevant_plugins = []
if not topic or scene_number is None:
logger.warning("Missing topic or scene_number for RAG error fix query generation")
return []
# Setup cache
cache_dir = self._create_cache_directory(topic, scene_number)
cache_file = cache_dir / "rag_queries_error_fix.json"
# Try to load from cache
cached_queries = self._load_cached_queries(cache_file)
if cached_queries is not None:
logger.info(f"Using cached RAG error fix queries for {topic}_scene{scene_number}")
return cached_queries
# Generate new queries for error fix
try:
plugins_text = ", ".join(relevant_plugins) if relevant_plugins else "No plugins are relevant."
prompt = get_prompt_rag_query_generation_fix_error(
error=error,
code=code,
relevant_plugins=plugins_text
)
response = self.helper_model(
_prepare_text_inputs(prompt),
metadata={
"generation_name": "rag-query-generation-fix-error",
"trace_id": scene_trace_id,
"tags": [topic, f"scene{scene_number}"],
"session_id": session_id
}
)
queries = self._extract_json_from_response(response, "RAG queries for error fix")
# Cache the queries
if queries:
self._save_queries_to_cache(queries, cache_file)
return queries
except Exception as e:
logger.error(f"Error generating RAG queries for error fix: {e}")
return []
def _extract_code_with_retries(
self,
response_text: str,
pattern: str = CODE_PATTERN,
generation_name: Optional[str] = None,
trace_id: Optional[str] = None,
session_id: Optional[str] = None,
max_retries: int = DEFAULT_MAX_RETRIES
) -> str:
"""Extract code from response text with retry logic.
Args:
response_text: The text containing code to extract
pattern: Regex pattern for extracting code
generation_name: Name of generation step
trace_id: Trace identifier
session_id: Session identifier
max_retries: Maximum number of retries
Returns:
The extracted code
Raises:
ValueError: If code extraction fails after max retries
"""
retry_prompt_template = """
Please extract the Python code in the correct format using the pattern: {pattern}.
You MUST NOT include any other text or comments.
You MUST return the exact same code as in the previous response, NO CONTENT EDITING is allowed.
Previous response:
{response_text}
"""
for attempt in range(max_retries):
try:
code_match = re.search(pattern, response_text, re.DOTALL)
if code_match:
extracted_code = code_match.group(1).strip()
logger.debug(f"Successfully extracted code on attempt {attempt + 1}")
return extracted_code
if attempt < max_retries - 1:
logger.warning(f"Attempt {attempt + 1}: Failed to extract code pattern. Retrying...")
# Regenerate response with a more explicit prompt
retry_prompt = retry_prompt_template.format(
pattern=pattern,
response_text=response_text[:1000] # Limit response length
)
response_text = self.scene_model(
_prepare_text_inputs(retry_prompt),
metadata={
"generation_name": f"{generation_name}_format_retry_{attempt + 1}",
"trace_id": trace_id,
"session_id": session_id
}
)
except Exception as e:
logger.error(f"Error during code extraction attempt {attempt + 1}: {e}")
if attempt == max_retries - 1:
break
raise ValueError(f"Failed to extract code pattern after {max_retries} attempts. Pattern: {pattern}")
def _prepare_additional_context(self, additional_context: Union[str, List[str], None]) -> List[str]:
"""Prepare additional context for code generation."""
if additional_context is None:
return []
elif isinstance(additional_context, str):
return [additional_context]
return additional_context.copy()
def _retrieve_rag_context(
self,
rag_queries: List[str],
scene_trace_id: Optional[str],
topic: str,
scene_number: int
) -> Optional[str]:
"""Retrieve context from RAG vector store."""
if not self.vector_store or not rag_queries:
return None
try:
return self.vector_store.find_relevant_docs(
queries=rag_queries,
k=DEFAULT_RAG_K_VALUE,
trace_id=scene_trace_id,
topic=topic,
scene_number=scene_number
)
except Exception as e:
logger.error(f"Error retrieving RAG context: {e}")
return None
def generate_manim_code(
self,
topic: str,
description: str,
scene_outline: str,
scene_implementation: str,
scene_number: int,
additional_context: Union[str, List[str], None] = None,
scene_trace_id: Optional[str] = None,
session_id: Optional[str] = None,
rag_queries_cache: Optional[Dict] = None
) -> Tuple[str, str]:
"""Generate Manim code from video plan.
Args:
topic: Topic of the scene
description: Description of the scene
scene_outline: Outline of the scene
scene_implementation: Implementation details
scene_number: Scene number
additional_context: Additional context
scene_trace_id: Trace identifier
session_id: Session identifier
rag_queries_cache: Cache for RAG queries (deprecated, use file cache)
Returns:
Tuple of generated code and response text
Raises:
ValueError: If code generation fails
"""
try:
# Prepare additional context
context_list = self._prepare_additional_context(additional_context)
# Add context learning examples if enabled
if self.use_context_learning and self.context_examples:
context_list.append(self.context_examples)
# Add RAG context if enabled
if self.use_rag:
rag_queries = self._generate_rag_queries_code(
implementation=scene_implementation,
scene_trace_id=scene_trace_id,
topic=topic,
scene_number=scene_number,
session_id=session_id or self.session_id
)
rag_context = self._retrieve_rag_context(
rag_queries, scene_trace_id, topic, scene_number
)
if rag_context:
context_list.append(rag_context)
# Generate prompt
prompt = get_prompt_code_generation(
scene_outline=scene_outline,
scene_implementation=scene_implementation,
topic=topic,
description=description,
scene_number=scene_number,
additional_context=context_list if context_list else None
)
# Generate code using model
response_text = self.scene_model(
_prepare_text_inputs(prompt),
metadata={
"generation_name": "code_generation",
"trace_id": scene_trace_id,
"tags": [topic, f"scene{scene_number}"],
"session_id": session_id or self.session_id
}
)
# Extract code with retries
code = self._extract_code_with_retries(
response_text,
CODE_PATTERN,
generation_name="code_generation",
trace_id=scene_trace_id,
session_id=session_id or self.session_id
)
logger.info(f"Successfully generated code for {topic} scene {scene_number}")
return code, response_text
except Exception as e:
logger.error(f"Error generating Manim code for {topic} scene {scene_number}: {e}")
raise ValueError(f"Code generation failed: {e}") from e
def fix_code_errors(
self,
implementation_plan: str,
code: str,
error: str,
scene_trace_id: str,
topic: str,
scene_number: int,
session_id: str,
rag_queries_cache: Optional[Dict] = None
) -> Tuple[str, str]:
"""Fix errors in generated Manim code.
Args:
implementation_plan: Original implementation plan
code: Code containing errors
error: Error message to fix
scene_trace_id: Trace identifier
topic: Topic of the scene
scene_number: Scene number
session_id: Session identifier
rag_queries_cache: Cache for RAG queries (deprecated, use file cache)
Returns:
Tuple of fixed code and response text
Raises:
ValueError: If code fixing fails
"""
try:
# Start with base error fix prompt
additional_context = None
# Add RAG context if enabled
if self.use_rag:
rag_queries = self._generate_rag_queries_error_fix(
error=error,
code=code,
scene_trace_id=scene_trace_id,
topic=topic,
scene_number=scene_number,
session_id=session_id
)
rag_context = self._retrieve_rag_context(
rag_queries, scene_trace_id, topic, scene_number
)
if rag_context:
additional_context = rag_context
# Generate prompt (with or without RAG context)
if additional_context:
prompt = get_prompt_fix_error(
implementation_plan=implementation_plan,
manim_code=code,
error=error,
additional_context=additional_context
)
else:
prompt = get_prompt_fix_error(
implementation_plan=implementation_plan,
manim_code=code,
error=error
)
# Get fixed code from model
response_text = self.scene_model(
_prepare_text_inputs(prompt),
metadata={
"generation_name": "code_fix_error",
"trace_id": scene_trace_id,
"tags": [topic, f"scene{scene_number}"],
"session_id": session_id
}
)
# Extract fixed code with retries
fixed_code = self._extract_code_with_retries(
response_text,
CODE_PATTERN,
generation_name="code_fix_error",
trace_id=scene_trace_id,
session_id=session_id
)
logger.info(f"Successfully fixed code errors for {topic} scene {scene_number}")
return fixed_code, response_text
except Exception as e:
logger.error(f"Error fixing code for {topic} scene {scene_number}: {e}")
raise ValueError(f"Code error fixing failed: {e}") from e
def visual_self_reflection(
self,
code: str,
media_path: Union[str, Image.Image],
scene_trace_id: str,
topic: str,
scene_number: int,
session_id: str
) -> Tuple[str, str]:
"""Use snapshot image or mp4 video to fix code.
Args:
code: Code to fix
media_path: Path to media file or PIL Image
scene_trace_id: Trace identifier
topic: Topic of the scene
scene_number: Scene number
session_id: Session identifier
Returns:
Tuple of fixed code and response text
Raises:
ValueError: If visual self-reflection fails
FileNotFoundError: If media file doesn't exist
"""
try:
# Validate media input
if isinstance(media_path, str):
media_file = Path(media_path)
if not media_file.exists():
raise FileNotFoundError(f"Media file not found: {media_path}")
# Determine if we're dealing with video or image
is_video = isinstance(media_path, str) and media_path.lower().endswith('.mp4')
# Load prompt template
prompt_file = Path('task_generator/prompts_raw/prompt_visual_self_reflection.txt')
if not prompt_file.exists():
logger.warning(f"Visual self-reflection prompt file not found: {prompt_file}")
# Fallback prompt
prompt_template = """
Analyze the visual output and the provided code. Fix any issues you notice in the code.
Code:
{code}
"""
else:
with prompt_file.open('r', encoding=CACHE_FILE_ENCODING) as f:
prompt_template = f.read()
# Format prompt
prompt = prompt_template.format(code=code)
# Prepare input based on media type and model capabilities
if is_video and isinstance(self.scene_model, (GeminiWrapper, VertexAIWrapper)):
# For video with Gemini models
messages = [
{"type": "text", "content": prompt},
{"type": "video", "content": str(media_path)}
]
else:
# For images or non-Gemini models
if isinstance(media_path, str):
media = Image.open(media_path)
else:
media = media_path
messages = [
{"type": "text", "content": prompt},
{"type": "image", "content": media}
]
# Get model response
response_text = self.scene_model(
messages,
metadata={
"generation_name": "visual_self_reflection",
"trace_id": scene_trace_id,
"tags": [topic, f"scene{scene_number}"],
"session_id": session_id
}
)
# Extract code with retries
fixed_code = self._extract_code_with_retries(
response_text,
CODE_PATTERN,
generation_name="visual_self_reflection",
trace_id=scene_trace_id,
session_id=session_id
)
logger.info(f"Successfully completed visual self-reflection for {topic} scene {scene_number}")
return fixed_code, response_text
except Exception as e:
logger.error(f"Error in visual self-reflection for {topic} scene {scene_number}: {e}")
raise ValueError(f"Visual self-reflection failed: {e}") from e
def enhanced_visual_self_reflection(
self,
code: str,
media_path: Union[str, Image.Image],
scene_trace_id: str,
topic: str,
scene_number: int,
session_id: str,
implementation_plan: Optional[str] = None
) -> Tuple[str, str]:
"""Enhanced visual self-reflection using VLM for detailed error detection.
This method specifically focuses on detecting and fixing:
- Element overlap and collision
- Out-of-bounds positioning
- Spatial boundary violations
- Poor visual arrangement
- Educational effectiveness issues
Args:
code: Code to analyze and fix
media_path: Path to media file or PIL Image
scene_trace_id: Trace identifier
topic: Topic of the scene
scene_number: Scene number
session_id: Session identifier
implementation_plan: Optional implementation plan for context
Returns:
Tuple of fixed code and response text
Raises:
ValueError: If enhanced visual analysis fails
FileNotFoundError: If media file doesn't exist
"""
try:
# Validate media input
if isinstance(media_path, str):
media_file = Path(media_path)
if not media_file.exists():
raise FileNotFoundError(f"Media file not found: {media_path}")
# Determine if we're dealing with video or image
is_video = isinstance(media_path, str) and media_path.lower().endswith('.mp4')
# Load enhanced visual analysis prompt
enhanced_prompt_file = Path('task_generator/prompts_raw/prompt_enhanced_visual_self_reflection.txt')
if enhanced_prompt_file.exists():
with enhanced_prompt_file.open('r', encoding=CACHE_FILE_ENCODING) as f:
prompt_template = f.read()
else:
# Fallback to original prompt if enhanced version not found
logger.warning("Enhanced visual self-reflection prompt not found, using fallback")
prompt_template = self._get_fallback_visual_prompt()
# Format prompt with implementation plan and code
prompt = prompt_template.format(
implementation=implementation_plan or "No implementation plan provided",
code=code
)
# Prepare input based on media type and model capabilities
if is_video and isinstance(self.scene_model, (GeminiWrapper, VertexAIWrapper)):
# For video with Gemini/Vertex AI models
messages = [
{"type": "text", "content": prompt},
{"type": "video", "content": str(media_path)}
]
else:
# For images or non-Gemini models
if isinstance(media_path, str):
media = Image.open(media_path)
else:
media = media_path
messages = [
{"type": "text", "content": prompt},
{"type": "image", "content": media}
]
# Get enhanced VLM analysis response
response_text = self.scene_model(
messages,
metadata={
"generation_name": "enhanced_visual_self_reflection",
"trace_id": scene_trace_id,
"tags": [topic, f"scene{scene_number}", "visual_error_detection"],
"session_id": session_id
}
)
# Parse response for visual analysis results
if "<LGTM>" in response_text or response_text.strip() == "<LGTM>":
logger.info(f"Enhanced visual analysis passed for {topic} scene {scene_number}")
return code, response_text
# Extract improved code if visual issues were found
fixed_code = self._extract_visual_fix_code(response_text, scene_trace_id, session_id)
logger.info(f"Enhanced visual self-reflection completed with fixes for {topic} scene {scene_number}")
return fixed_code, response_text
except Exception as e:
logger.error(f"Error in enhanced visual self-reflection for {topic} scene {scene_number}: {e}")
# Fallback to original visual_self_reflection if enhanced version fails
logger.info("Falling back to original visual_self_reflection method")
return self.visual_self_reflection(
code, media_path, scene_trace_id, topic, scene_number, session_id
)
def _extract_visual_fix_code(
self,
response_text: str,
scene_trace_id: Optional[str] = None,
session_id: Optional[str] = None
) -> str:
"""Extract code from enhanced visual analysis response.
Args:
response_text: The VLM response containing visual analysis
scene_trace_id: Trace identifier
session_id: Session identifier
Returns:
The extracted and fixed code
Raises:
ValueError: If code extraction fails
"""
# Try to extract code from <improved_code> tags first
improved_code_pattern = r'<improved_code>\s*```python\s*(.*?)\s*```\s*</improved_code>'
code_match = re.search(improved_code_pattern, response_text, re.DOTALL)
if code_match:
extracted_code = code_match.group(1).strip()
logger.debug("Successfully extracted code from <improved_code> tags")
return extracted_code
# Fallback to standard code extraction
return self._extract_code_with_retries(
response_text,
CODE_PATTERN,
generation_name="enhanced_visual_fix",
trace_id=scene_trace_id,
session_id=session_id
)
def _get_fallback_visual_prompt(self) -> str:
"""Get fallback visual analysis prompt if enhanced version is not available."""
return """
Analyze the visual output and the provided code for the following issues:
1. **Element Overlap:** Check for overlapping text, shapes, or mathematical expressions
2. **Out-of-Bounds Objects:** Identify elements outside the visible frame
3. **Spacing Issues:** Verify minimum 0.3 unit spacing between elements
4. **Safe Area Compliance:** Ensure 0.5 unit margins from frame edges
5. **Educational Clarity:** Assess if arrangement supports learning objectives
Implementation Plan: {implementation}
Code to analyze:
{code}
If issues are found, provide fixed code. If no issues, return "<LGTM>".
<improved_code>
```python
[Fixed code here]
```
</improved_code>
"""
def detect_visual_errors(
self,
media_path: Union[str, Image.Image],
scene_trace_id: Optional[str] = None,
topic: Optional[str] = None,
scene_number: Optional[int] = None,
session_id: Optional[str] = None
) -> Dict[str, Any]:
"""Detect visual errors using VLM without code modification.
This method provides detailed visual error analysis without attempting to fix code.
Useful for validation and quality assessment.
Args:
media_path: Path to media file or PIL Image
scene_trace_id: Trace identifier
topic: Topic of the scene
scene_number: Scene number
session_id: Session identifier
Returns:
Dictionary containing visual error analysis results
Raises:
ValueError: If visual error detection fails
FileNotFoundError: If media file doesn't exist
"""
try:
# Validate media input
if isinstance(media_path, str):
media_file = Path(media_path)
if not media_file.exists():
raise FileNotFoundError(f"Media file not found: {media_path}")
# Create analysis prompt
analysis_prompt = """
You are an expert visual quality analyst. Analyze this Manim-generated frame/video for:
1. **Element Overlap Detection:**
- Text overlapping with shapes or other text
- Mathematical expressions colliding
- Unintentional object occlusion
2. **Spatial Boundary Issues:**
- Objects extending beyond frame boundaries
- Violations of safe area margins (0.5 units from edges)
- Insufficient spacing between elements (minimum 0.3 units)
3. **Visual Quality Assessment:**
- Overall composition balance
- Readability of text elements
- Educational effectiveness of arrangement
Provide your analysis in the following format:
**VISUAL ERROR ANALYSIS:**
- Overlap Issues: [List any overlapping elements]
- Boundary Violations: [List out-of-bounds elements]
- Spacing Problems: [List spacing violations]
- Quality Issues: [List other visual problems]
**SEVERITY ASSESSMENT:**
- Critical Errors: [Issues that severely impact readability]
- Major Errors: [Issues that noticeably reduce quality]
- Minor Errors: [Issues that slightly affect visual appeal]
**OVERALL RATING:** [Excellent/Good/Fair/Poor]
"""
# Determine media type and prepare input
is_video = isinstance(media_path, str) and media_path.lower().endswith('.mp4')
if is_video and isinstance(self.scene_model, (GeminiWrapper, VertexAIWrapper)):
messages = [
{"type": "text", "content": analysis_prompt},
{"type": "video", "content": str(media_path)}
]
else:
if isinstance(media_path, str):
media = Image.open(media_path)
else:
media = media_path
messages = [
{"type": "text", "content": analysis_prompt},
{"type": "image", "content": media}
]
# Get analysis response
response_text = self.scene_model(
messages,
metadata={
"generation_name": "visual_error_detection",
"trace_id": scene_trace_id,
"tags": [topic or "unknown", f"scene{scene_number or 0}", "quality_analysis"],
"session_id": session_id or self.session_id
}
)
# Parse response into structured results
analysis_results = self._parse_visual_analysis(response_text)
logger.info(f"Visual error detection completed for scene {scene_number or 'unknown'}")
return analysis_results
except Exception as e:
logger.error(f"Error in visual error detection: {e}")
raise ValueError(f"Visual error detection failed: {e}") from e
def _parse_visual_analysis(self, response_text: str) -> Dict[str, Any]:
"""Parse visual analysis response into structured data.
Args:
response_text: Raw response from VLM
Returns:
Structured analysis results
"""
results = {
"overlap_issues": [],
"boundary_violations": [],
"spacing_problems": [],
"quality_issues": [],
"critical_errors": [],
"major_errors": [],
"minor_errors": [],
"overall_rating": "Unknown",
"raw_analysis": response_text
}
try:
# Extract different sections using regex patterns
overlap_match = re.search(r'Overlap Issues:\s*(.*?)(?=\n-|\n\*\*|$)', response_text, re.DOTALL)
if overlap_match:
results["overlap_issues"] = [item.strip() for item in overlap_match.group(1).split('\n') if item.strip()]
boundary_match = re.search(r'Boundary Violations:\s*(.*?)(?=\n-|\n\*\*|$)', response_text, re.DOTALL)
if boundary_match:
results["boundary_violations"] = [item.strip() for item in boundary_match.group(1).split('\n') if item.strip()]
rating_match = re.search(r'OVERALL RATING.*?:\s*([A-Za-z]+)', response_text)
if rating_match:
results["overall_rating"] = rating_match.group(1)
except Exception as e:
logger.warning(f"Error parsing visual analysis: {e}")
return results |