All Classes and Interfaces
Class
Description
Abstract base class for all agents in the graph system.
A2A Agent Card Provider.
The Wrapper of AgentCard.
Factory class for creating MethodToolCallback instances for ReactAgent.
Executor class for AgentTool that contains the method to be used by MethodToolCallback.
JSON array loop strategy that retrieves a JSON array from the current message state,
sends each array element as a message to the model, and returns the result.
Strategy for building conditional execution graphs.
A node action that evaluates conditions to determine the next execution path.
An async edge action that evaluates conditions to determine routing paths.
Conditional loop strategy that retries until the Predicate is satisfied or the maximum count is reached.
Context editing interceptor that clears older tool results once the conversation
grows beyond a configurable token threshold.
Fixed count loop strategy
Tool for editing files using string replacement.
Request structure for editing a file.
Result of edit operation.
Enhanced parallel result aggregator that supports custom merge strategies and
concurrency control.
FileData structure for storing file contents with metadata.
Structured file listing information.
Backend interface for file storage operations.
Filesystem interceptor that provides file system management capabilities to agents.
Builder for FilesystemInterceptor with comprehensive configuration options.
Abstract base builder for FlowAgent and its subclasses.
A centralized builder for constructing StateGraphs for different FlowAgent types.
Configuration class for graph building parameters.
Strategy interface for building StateGraphs for different FlowAgent types.
Registry for FlowGraphStrategy implementations.
Glob search tool for fast file pattern matching.
Tool for finding files matching a glob pattern.
Structured grep match entry.
Grep search tool for fast content search.
Tool for searching text patterns in files.
Request structure for grep search.
Enum representing the position where a hook should be executed in the agent workflow.
Annotation to specify the position(s) where a hook should be executed.
Utility class for chaining model and tool interceptors.
Hook that supports interruption and feedback mechanism.
Enumeration representing jump destinations in the agent workflow.
Tool interceptor that automatically evicts large tool results to the filesystem.
Builder for LargeResultEvictionInterceptor.
Tool for listing files in a directory.
Builder for creating LlmRoutingAgent instances.
Backend that reads and writes files directly from the filesystem.
Loop Agent that supports multiple loop modes:
COUNT: Execute a fixed number of loops
CONDITION: Continue looping based on a condition, similar to a do-while
structure, but when the condition is true, terminate the loop
JSON_ARRAY: Parse a JSON array and iterate over its elements
Other Loop Strategy: Users can implement the LoopStrategy interface according to their needs.
Converts a LoopAgent into its corresponding StateGraph.
Built-in loop strategies for LoopAgent
Loop strategy for LoopAgent, used to control the behavior of LoopAgent.
Internal static class that proxies MessagesAgentHook and implements
AsyncNodeActionWithConfig interface for afterAgent hook.
Internal static class that proxies MessagesAgentHook and implements
AsyncNodeActionWithConfig interface.
Internal static class that proxies MessagesModelHook and implements
AsyncNodeActionWithConfig interface for afterModel hook.
Internal static class that proxies MessagesModelHook and implements
AsyncNodeActionWithConfig interface.
Handler interface for model call interceptors.
Exception raised when model call limits are exceeded.
Hook that tracks and limits model call counts.
Automatic fallback to alternative models on errors.
Model interceptor that can wrap model calls.
Request object for model calls.
Response object for model calls.
ParallelAgent executes multiple sub-agents in parallel and merges their results.
Concatenation merge strategy for string results.
Default merge strategy that combines all results into a map.
List merge strategy that combines results into a list.
Strategy interface for merging parallel execution results.
Builder for creating ParallelAgent instances.
Strategy for building parallel execution graphs.
A node action that aggregates results from parallel execution of multiple agents.
Middleware to patch dangling tool calls in the messages history.
Builder for creating PatchToolCallsInterceptor instances.
Exception raised when PII is detected with BLOCK strategy.
Detect and handle Personally Identifiable Information (PII) in agent conversations.
Functional interface for detecting PII in text.
Built-in PII detectors for common types.
Represents a detected instance of PII in text.
Types of Personally Identifiable Information that can be detected.
Interface for objects that can be prioritized and sorted by their order value.
Tool for reading file contents with pagination support.
Request structure for reading a file.
Strategies for handling detected PII.
Response record for structured routing decision output
Strategy for building LLM-based routing graphs.
Builder for creating SequentialAgent instances.
Strategy for building sequential execution graphs.
Manages shell sessions and command execution.
Result of command execution.
Simple pattern-based redaction rule.
Result of redaction operation with match information.
Redaction rule for sanitizing command output.
A tool for executing shell commands.
Defines the parameters for a shell tool request.
Hook for managing ShellTool session.
Builder class for constructing ShellToolAgentHook instances.
SubAgent interceptor that provides subagent invocation capabilities to agents.
Specification for creating a subagent.
Hook that summarizes conversation history when token limits are approached.
Builder for creating SupervisorAgent instances.
Response record for structured supervisor decision output.
Strategy for building supervisor-based routing graphs.
Tool that enables invoking subagents to handle complex, isolated tasks.
Request structure for the task tool.
Model interceptor that provides todo list management capabilities to agents.
Represents a single todo item.
Todo item status.
Functional interface for counting tokens in messages.
Handler interface for tool call interceptors.
Exception raised when tool call limits are exceeded.
Hook that tracks and limits tool call counts.
Request object for tool calls.
Response object for tool calls.
Tool interceptor that emulates specified tools using an LLM instead of executing them.
Interface for hooks that need a specific tool injected.
Tool interceptor that can wrap tool calls.
Tool interceptor that automatically retries failed tool calls with configurable backoff.
Uses an LLM to select relevant tools before calling the main model.
Enumeration for message append policy.
Tool for writing content to a new file.
Request structure for writing a file.
Result of write operation.
Tool for writing and managing todos in the agent workflow.