Framework Namespace. More...

Namespaces
namespace	llms
	Large Language Models Namespace.

namespace	media
	Media envelope namespace.

namespace	tools
	Tools Namespace.

namespace	workflows
	Worflows Namespace.

Classes
class	ActorAgent
	Actor-based agent implementation. More...

class	Agent
	Interface for agents. More...

class	AgentContext
	Context for an agent, containing tools, LLM, and memory. More...

class	AsyncGenerator
	AsyncGenerator with a Folly-compatible API: next() returns Task<optional<T>> More...

struct	AudioFormat
	Audio format specification for live AI sessions. More...

class	AutonomousAgent
	An agent that operates autonomously to complete a task. More...

class	ConfigLoader
	A utility class to load and access configuration values from .env files. More...

class	Executor
	A minimal executor implementation (fire-and-forget) More...

class	ILiveClient
	Abstract interface for live real-time AI clients. More...

struct	LiveSessionConfig
	Configuration for live AI sessions. More...

class	LLMInterface
	Interface for language model providers (OpenAI, Anthropic, Google, Ollama) More...

struct	LLMOptions
	Options for LLM API calls. More...

struct	LLMResponse
	Response from an LLM. More...

class	Logger
	Logger utility class that wraps spdlog functionality. More...

class	Memory
	Interface for agent memory storage. More...

struct	Message
	Message in a conversation. More...

struct	Parameter
	Parameter type for tools and LLM calls. More...

class	Task
	Standard C++20 coroutine-based Task implementation (no external deps) More...

class	Task< void >
	Task specialization for void. More...

class	Tool
	Interface for tools that an agent can use. More...

struct	ToolResult
	Result of a tool execution. More...

Typedefs
using	ToolCallback = std::function<ToolResult(const JsonObject&)>
	Callback type for tool execution.

using	String = std::string
	String type.

using	StringMap = std::map<String, String>
	String map type.

using	JsonObject = nlohmann::json
	JSON object type.

using	JsonArray = nlohmann::json::array_t
	JSON array type.

using	ParameterMap = std::map<String, Parameter>
	Parameter map type.

Enumerations
enum class	TransportProtocol { WEBSOCKET , WEBRTC , HTTP_STREAM , CUSTOM }
	Transport protocol types for live AI communication. More...

enum class	MemoryType { SHORT_TERM , LONG_TERM , WORKING }
	Memory types. More...

Functions
template<typename T>
T	blockingWait (Task< T > &&task)
	Helper to run a coroutine and get the result synchronously.

void	blockingWait (Task< void > &&task)
	Helper to run a coroutine and get the result synchronously.

template<typename T>
Task< std::vector< T > >	collectAllFromGenerator (AsyncGenerator< T > &&generator)
	Helper to collect all results from an AsyncGenerator.

template<typename T>
std::vector< T >	collectAll (AsyncGenerator< T > &&generator)
	Helper to run an async generator and collect results.

Executor *	getExecutor ()
	Get a global executor.

std::shared_ptr< LLMInterface >	createLLM (const String &provider, const String &api_key, const String &model="")
	Factory function to create a specific LLM provider.

std::shared_ptr< Memory >	createMemory ()
	Create a Memory instance.

std::shared_ptr< Tool >	createTool (const String &name, const String &description, const std::vector< Parameter > &parameters, ToolCallback callback)
	Create a custom tool with a name, description, parameters, and callback.

Detailed Description

Framework Namespace.

Typedef Documentation

◆ JsonArray

using agents::JsonArray = nlohmann::json::array_t

JSON array type.

Note: This is a JSON array type. It contains the JSON array.

◆ JsonObject

using agents::JsonObject = nlohmann::json

JSON object type.

Note: This is a JSON object type. It contains the JSON object.

◆ ParameterMap

using agents::ParameterMap = std::map<String, Parameter>

Parameter map type.

Note: This is a parameter map type. It contains the string and parameter.

◆ String

using agents::String = std::string

String type.

Note: This is a string type. It contains the string.

◆ StringMap

using agents::StringMap = std::map<String, String>

String map type.

Note: This is a string map type. It contains the string and string.

◆ ToolCallback

using agents::ToolCallback = std::function<ToolResult(const JsonObject&)>

Callback type for tool execution.

Note: This is the callback type for tool execution. It is a function that takes a json object and returns a tool result.

Enumeration Type Documentation

◆ MemoryType

enum class agents::MemoryType

strong

Memory types.

Note: This is a type of memory. It contains the type of memory, the name of the memory, and the content of the memory.

Enumerator
SHORT_TERM	Short term memory.
LONG_TERM	Long term memory.
WORKING	Working memory.

◆ TransportProtocol

enum class agents::TransportProtocol

strong

Transport protocol types for live AI communication.

Different protocols offer different trade-offs in terms of latency, reliability, and browser compatibility.

Enumerator
WEBSOCKET	WebSocket-based communication (high compatibility, medium latency)
WEBRTC	WebRTC-based communication (low latency, requires ICE servers)
HTTP_STREAM	HTTP streaming (fallback option, high latency)
CUSTOM	Custom protocol implementation.

Function Documentation

◆ blockingWait() [1/2]

template<typename T>

T agents::blockingWait ( Task< T > && task )

Helper to run a coroutine and get the result synchronously.

Template Parameters

T	The type of the result

Parameters

task	The task to run

Returns: The result of the task

◆ blockingWait() [2/2]

void agents::blockingWait ( Task< void > && task )

inline

Helper to run a coroutine and get the result synchronously.

Parameters

task	The task to run

◆ collectAll()

template<typename T>

std::vector< T > agents::collectAll ( AsyncGenerator< T > && generator )

Helper to run an async generator and collect results.

Template Parameters

T	The type of the result

Parameters

generator The generator to collect results from

Returns: The results of the generator

◆ collectAllFromGenerator()

template<typename T>

Task< std::vector< T > > agents::collectAllFromGenerator ( AsyncGenerator< T > && generator )

Helper to collect all results from an AsyncGenerator.

Template Parameters

T	The type of the result

Parameters

generator The generator to collect results from

Returns: The results of the generator

◆ createLLM()

std::shared_ptr< LLMInterface > agents::createLLM	(	const String &	provider,
		const String &	api_key,
		const String &	model = "" )

Factory function to create a specific LLM provider.

Parameters

provider	One of: "anthropic", "openai", "google", "ollama"
api_key	API key for the provider
model	Model to use (provider-specific)

Returns: The LLMInterface

◆ createMemory()

std::shared_ptr< Memory > agents::createMemory ( )

Create a Memory instance.

Returns: The Memory instance

◆ createTool()

std::shared_ptr< Tool > agents::createTool	(	const String &	name,
		const String &	description,
		const std::vector< Parameter > &	parameters,
		ToolCallback	callback )

Create a custom tool with a name, description, parameters, and callback.

Parameters

name	The name of the tool
description	The description of the tool
parameters	The parameters of the tool
callback	The callback for the tool

Returns: A shared pointer to the tool

◆ getExecutor()

Executor * agents::getExecutor ( )

inline

Get a global executor.

Returns: Pointer to global executor

Namespaces

Classes

Typedefs

Enumerations

Functions

Detailed Description

Typedef Documentation

◆ JsonArray

◆ JsonObject

◆ ParameterMap

◆ String

◆ StringMap

◆ ToolCallback

Enumeration Type Documentation

◆ MemoryType

◆ TransportProtocol

Function Documentation

◆ blockingWait() [1/2]

◆ blockingWait() [2/2]

◆ collectAll()

◆ collectAllFromGenerator()

◆ createLLM()

◆ createMemory()

◆ createTool()

◆ getExecutor()