Converts a list of text strings into the input format for the Agent model. Args: texts (List[str]): The list of text strings to be p

Load context learning examples of a specific type from files. Args: example_type (str): Type of examples to load ('scene_plan', '

Combine all videos and subtitle files for a specific topic using VideoRenderer. Args: topic (str): The topic to combine

Extract XML content from a text response. Extracts XML content between ```xml markers. Returns the full response if no XML blocks found. Arg

Creates an embedding function using litellm. Returns: Embeddings: A LangChain Embeddings instance that wraps litellm functionalit

Extract code from response text with retry logic. Args: response_text (str): The text containing code to extract patt

Calculate the geometric mean of a list of scores. Args: scores (List[int]): List of integer scores, may contain None values. Re

Convert score fields in a dictionary to integers recursively. Args: data (dict): Dictionary containing score fields to convert.

Extract JSON content from a string response. Args: response (str): String containing JSON content, possibly within code blocks.

Finds relevant documentation based on the provided queries. Args: queries (List[Dict]): List of query dictionaries with 'type' an

For generating RAG queries during technical implementation to code generation stage

Get relevant documentation using the vector store. Args: rag_queries (List[Dict]): List of RAG queries to search for

Recursively merge two dictionaries. Args: dict1 (dict): First dictionary. dict2 (dict): Second dictionary. Returns:

Process a topic name by capitalizing words and handling special characters. Args: topic_name (str): The topic name to process.

Render the video for all scenes with code fixing capability. Args: topic (str): The topic of the video descr

Save individual evaluation result to a file. Args: output_folder (str): Directory to store the evaluation file. file_name (s

Adds documents to a vector store in batches with rate limiting. Args: vector_store (Chroma): The vector store to add documents to

Generate implementation plan for a single scene. Args: topic (str): The topic of the video description (str): Descrip

Get the MIME type of a file based on its extension Args: file_path: Path to the file Return

Load existing session ID from file or create a new one. Returns: str: The session ID either loaded from file or newly cr

Processes documentation files from a folder into LangChain documents. Args: folder_path (str): Path to the folder containing docu

Detect which plugins might be relevant based on topic and description. Args: topic (str): Topic of the video descript

Evaluate sampled frames from a video using an image evaluation model. Args: model: The image evaluation model to use video_path (

Evaluate a text file using the provided model. Args: model: The model to use for evaluation. transcript_path (str): Path to

Evaluate a video file using the provided model. Args: model: The model to use for evaluation. video_path (str): Path to the

Modified pipeline to handle partial scene completions and option to only generate plans for specific scenes. Args: topic

Generate a prompt to fix errors in the given manim code. Args: implementation_plan (str): The implementation plan of the scene.

Find and save the image with the most non-black space from a list of images. Args: images (list): List of image file paths, PIL Image obj

Process a theorem file or directory for evaluation. Args: models: Dictionary of models for different evaluation types. file_

Set the relevant plugins for the current video. Args: plugins (List[str]): List of plugin names to set as relevant

Creates the main ChromaDB vector store for Manim core documentation. Returns: Chroma: The initialized and populated core vector s

Download a file from a URL and save it as a temporary file Args: url: URL of the file to download

Encode local file or PIL Image to base64 string Args: file_path: Path to local file or PIL Image object

Format examples using the appropriate template based on their type. Args: example_type (str): Type of examples to format

Generate RAG queries from the implementation plan. Args: implementation (str): The implementation plan text scene_tra

Generate RAG queries for fixing code errors. Args: error (str): The error message to fix code (str): The code contain

Generate RAG queries from the storyboard to help create narration plan. Args: storyboard (str): Storyboard text to generate queri

Generate RAG queries from the scene plan to help create storyboard. Args: scene_plan (str): Scene plan text to generate queries f

Generate RAG queries from the storyboard to help create technical implementation. Args: storyboard (str): Storyboard text to gene

Generate detailed implementation plan for a single scene using VideoPlanner. Args: topic (str): The topic of the video

Get the MIME type of a file based on its extension Args: file_path: Path to the file Return

Load all context learning examples from the specified directory. Returns: str: Formatted context learning examples, or None if no

Loads existing or creates new ChromaDB vector stores. Creates/loads vector stores for both Manim core documentation and any available plugins

Load plugin descriptions from JSON file. Returns: list: List of plugin descriptions, empty list if loading fails

Converts text strings and images into the input format for the Agent model. Args: texts (Union[str, List[str]]): Text string(s) to b

Converts text strings and video file paths into the input format for the Agent model. Args: texts (Union[str, List[str]]): Text stri

Save a PIL Image to a temporary file Args: image: PIL Image object Returns: Pat

Save session ID for a specific topic. Args: topic (str): The topic to save the session ID for session_id (st

Uploads the given file to Gemini. Args: file_path: Path to the file mime_type: MIME type of the file

Calculate the overall score from evaluation results. Args: result (Dict): Dictionary containing evaluation results. Returns:

Find the prompts_raw directory and generate an __init__.py file containing prompt texts. Searches for prompts_raw directory in current and p

Locates the prompts_raw directory and generates an __init__.py file containing prompt texts. Searches for prompts_raw directory in current a

Check if a theorem has its plan, code files, and rendered videos with detailed scene status. Args: theorem (Dict): Dicti

Combine all evaluation results into a single file. Args: output_folder (str): Directory to store the combined file. combined

Generate a Python file containing prompt texts from .txt files. Args: folder_path (str): Path to directory containing prompt .txt fi

Creates a Python file containing prompt texts from .txt files. Args: folder_path (str): Path to directory containing prompt .txt fil

Create a snapshot of the video for a specific topic and scene. Args: topic (str): Topic name scene_number (int): Scen

Evaluate transcript text using an LLM model with retry logic. Args: text_eval_model: The LLM model wrapper to use for evaluation.

Evaluate a single video chunk using a multimodal model. Args: model: The multimodal model to use for evaluation video_path (

Extract key frames from a video by dividing it into chunks and selecting representative frames. Args: video_path (str): Path to the input

Extract all score values from a nested dictionary or list structure. Args: data (Union[Dict, List]): The data structure to extract s

Fix errors in generated Manim code. Args: implementation_plan (str): Original implementation plan code (str): Code co

Fix and clean up a transcript using an LLM model. Args: text_eval_model: The LLM model wrapper to use for fixing the transcript.

Generate Manim code from video plan. Args: topic (str): Topic of the scene description (str): Description of the scen

Generate scene outline using VideoPlanner. Args: topic (str): The topic of the video description (str): Desc

Generates a hash based on the input data dictionary. The hash is used to create a unique identifier for the input data. Para

Extract frames from a video file at specified FPS. Args: video_path (str): Path to the video file. fps (float, optional): Frames

Generate a prompt for code generation based on the given video plan and implementation details. Args: topic (str): The topic of the

Generate a prompt for detecting relevant plugins based on topic and description. Args: topic (str): The video topic descript

For generating RAG queries during storyboard to narration stage

For generating RAG queries during storyboard to technical implementation stage

Generate a prompt for scene planning based on the given parameters. Args: topic (str): The topic of the video. description (

Load implementation plans for each scene. Args: topic (str): The topic to load implementation plans for Returns

Main function to run the evaluation script. Parses command line arguments and orchestrates the evaluation process for text, video, and i

Extract frames from video at subtitle timestamps and save with corresponding text. Args: output_dir (str): Directory containing the topic

Convert SRT subtitle file to plain text. Args: output_dir (str): Directory containing the topic folders. topic_name (str): Name o

Parse an SRT subtitle file into plain text. Args: srt_path: Path to the SRT subtitle file. Returns: str: The subtitle t

Process all topic folders in the output directory. Args: output_folder (str): Directory containing the topic folders.

Process a single scene using CodeGenerator and VideoRenderer. Args: i (int): Scene index scene_outline (str)

Reduces the frame rate of a video by only keeping frames at the target interval. Args: input_path (str): Path to the input video

Render a single scene and handle error retries and visual fixes. Args: code (str): The Manim code to render file_pref

Process messages and return completion Args: messages: List of message dictionaries with 'type' and 'content' ke

Process messages and return completion Args: messages: List of message dictionaries with 'type' and 'content' ke

Process messages and return completion. Args: messages: List of message dictionaries containing type and content

Initialize the Gemini wrapper Args: model_name: Name of the model to use temperature: Temperature fo

Initialize the LiteLLM wrapper Args: model_name: Name of the model to use (e.g. "azure/gpt-4", "vertex_ai/gemini

Initialize the Vertex AI wrapper. Args: model_name: Name of the model to use (e.g. "gemini-1.5-pro") temperat

Initialize the VideoRenderer. Args: output_dir (str, optional): Directory for output files. Defaults to "output". pri