SVG diagram builder.

Standard response format for all perception tool actions.

AI Agent with configurable LLM providers and context modes for ablation studies

Pure Python grep implementation

Executes bash commands in persistent shell sessions

Implementation of Elo rating system for model comparison. The Elo system updates ratings based on pairwise comparison outcomes, wher

Helper class for LLM-based operations.

A text-based game with hidden mechanics that agents must discover: 1. Certain colored keys open corresponding colored doors 2. Guards blo

Main agent class supporting multiple extraction modes

Container for multimodal content

Reads files from the local filesystem

Performs exact string replacements in files

Lists files and directories

Completely replaces the contents of a specific cell in a Jupyter notebook

Configuration for system hints

Writes files to the local filesystem

A chunk of conversation with metadata

Single message in a conversation

Fast file pattern matching tool

ReAct Agent with different KV cache optimization modes

Indexes and searches conversation chunks using the retrieval pipeline service

Indexes and searches conversation chunks using the retrieval pipeline service

Makes multiple edits to a single file in one operation

AI Agent with enhanced system hints for better trajectory management

An agent that learns from experience and can replay learned workflows. This agent: 1. Attempts to match tasks to learned workflows

Creates and manages structured task lists

Registry for managing available tools

Agentic RAG system with ReAct pattern and multiple LLM provider support

A chunk of conversation with metadata

Generic execution tools with safety and result analysis.

Configuration for the background memory processor

Unified message format

Represents an advanced memory card with complete metadata. Each card MUST include: - backstory: Context about when/why this informat

Background processor that analyzes conversations and updates memory Runs separately from the main conversation flow

Represents compressed content

Single message in a conversation

Pure conversational agent that focuses on dialogue Reads memory for context but doesn't update it directly

External system integration tools.

File system tools with verification and safety checks.

GPT-5 Agent with Native Tool Support This agent uses GPT-5's native web_search and code_interpreter capabilities through the OpenRou

Tools for searching and retrieving user memory information

高级 Web Search Agent - 扩展功能

Main coding agent with streaming support and modular tool system

Main configuration class for multimodal agent

Evaluates the dual memory system on test cases

Dual-layer memory indexer combining: 1. Contextual RAG for conversation chunks 2. Advanced JSON cards for structured facts

Agent with dual memory system: structured cards + contextual RAG

Configuration for the conversational agent

Manages storage and retrieval of learned workflows. The knowledge base provides: - Persistent storage of workflows - Intent matc

LLM-based evaluator for agent responses.

Local implementations of file system tools

Tools for searching and retrieving user memory information

Main search engine combining all components

Configuration for system hints

System state tracking for system hints

Registry for available tools

Evaluates the RAG agent on user memory test cases

Represents a complete workflow that can be learned and replayed

Configuration for the System-Hint Agent

Retrieves output from running or completed background bash shells

Enhanced chunk with contextual information

Implements contextual chunking inspired by Anthropic's Contextual Retrieval. Key Educational Points: 1. Context Generation: Uses LLM

Chunks conversation histories into segments for RAG indexing

A conversation history containing multiple messages.

Document data class.

Event-Triggered AI Agent with System Hints Responds to events while maintaining all system hint capabilities

GraphRAG knowledge graph indexer with entity extraction and community detection.

Inverted index data structure for efficient term lookup

Knowledge base backend types

Knowledge base backend types

LLM-based agent that uses in-context learning to play the game. Stores experiences and uses them to reason about future actions.

Agent using Ollama's native tool calling support

Metadata for PubChem operation results.

Q-learning agent for the treasure hunt game. Uses tabular Q-learning with state-action pairs.

RAPTOR tree-based document indexer with recursive summarization.

AI Agent for researching with context compression

Search result from knowledge base

Search result from knowledge base

Universal tool calling agent that works on all platforms Automatically selects vLLM (if GPU available) or Ollama

Agent that uses RAG to answer questions about user conversation history

Agent that uses RAG to answer questions about user conversation history

Web Search Agent - 使用 Kimi API 的内置搜索工具 根据官方文档: https://platform.moonshot.ai/docs/guide/use-web-search Kimi 提供了内置的 $web_search 工具

Metadata for Yahoo Finance operation results.

Agentic RAG system with ReAct pattern and multiple LLM provider support

Document chunking configuration

Configuration for conversation chunking

Represents a community of related entities.

Enhanced knowledge base tools with contextual retrieval support. Key Educational Points: 1. Dual Indexing: Maintains both contextual

Manages conversation history with optional vector search

Result of evaluating an agent's response.

Use this tool when you are in plan mode and ready to code

Represents a single game interaction experience.

Configuration for RAG indexing

Indexing modes for conversation chunks