3. Data, Science & AI

Generates a zoomable HTML visualization of spatial maps displaying walkability, hazards, resources, and observation confidence.

Loads persistent notes or collections by unique ID or name from storage systems.

Performs boolean text checks against natural language predicates with automatic chunking for long content, returning true on any matching segment.

Enumerates possible navigation positions within a bounded step count using over-approximation for simulation purposes.

Updates persistent spatial maps by integrating observation data into cell-based grid structures for real-time environmental mapping.

Extracts a specified field value from every note in a collection, enabling efficient data aggregation and transformation.

Initializes a ScienceWorld simulation episode by reading scenario, difficulty, and seed from YAML configuration, returning initial observation and session ID for agent interaction.

Creates interactive 3D HTML visualization of session-local occupancy grid data for technical inspection and analysis.

Refines unstructured or semi-structured text by extracting or transforming existing information using an LLM, without adding new content.

Enables/disables automatic capture of training data including partial observation grid and optional ground truth for AI model development.

Enables AI agents to drop inventory items as world entities in Minecraft, returning the dropped items for simulation tasks.

Queries the agent's persistent spatial map for environmental knowledge, enabling reasoning, explanation, and planning based on integrated observations.

Restores autonomous systems to a previously safe navigation state through controlled, limited movement to avoid obstacles.

Validates SELECT field implementation for accurate data retrieval in a project's data layer.

Retrieves current OSWorld environment observation: base64 screenshot and accessibility tree JSON for data processing.

Checks if a given text contains a question using pattern matching for question marks or interrogative structure.

Automates agent movement to achieve stable adjacency with a vertical surface using navigation conventions and motion tools.

Evaluates mathematical expressions numerically using the SymPy library for symbolic mathematics.

Enables deep learning model training, GPU acceleration, and end-to-end data science workflows using PyTorch.

Filters and analyzes stocks using financial metrics from CSV data, enabling comparison and screening based on key indicators.

Retrieves real-time and historical financial data for Chinese markets including A-shares, Hong Kong, US, indices, funds, and futures.

Transforms legacy codebases into AI-ready projects by generating Claude Code configurations, including CLAUDE.md and AI-assisted development standards.

Interprets Google Analytics 4 and Search Console data, delivering benchmarks, status indicators, and actionable business insights.

Provides efficient, readable, and performant data manipulation standards for Python's Pandas library.