Crate gestura_core

Expand description

Gestura’s public core facade.

gestura-core is the stable Rust API surface for Gestura.app. It owns the cross-domain orchestration that ties the workspace together and re-exports focused gestura-core-* domain crates so downstream code can import through a single, consistent path.

§Generated docs quick start

If you are exploring the generated docs, start with these high-signal module entry points:

pipeline: main agent execution flow and request/response model
tools: built-in tools, permissions, schemas, and confirmation helpers
config: application configuration plus secret-hydration/runtime bridges
llm_provider: provider selection and LLM facade access
mcp and a2a: protocol surfaces for MCP and agent-to-agent flows
agent_sessions, session_manager, and tasks: session and task state
knowledge, memory_bank, and memory_console: expert knowledge and memory
security and streaming: security/runtime streaming boundaries

A good reading path is:

start here in gestura-core
jump into one of the facade modules above
follow the re-export or module docs into the owning gestura-core-* crate when you need domain-level detail

§Core-first architecture

The workspace follows a core-first layout:

domain logic lives in dedicated crates such as gestura-core-tools, gestura-core-mcp, gestura-core-config, and gestura-core-context
gestura-core re-exports those crates under stable public module names
presentation layers (gestura-cli, gestura-gui) stay thin and delegate to this facade instead of owning business logic

In practice, this means new work usually follows this path:

add or update behavior in the relevant domain crate
expose the stable public entry point from gestura-core
consume the stable API from CLI and GUI code

§What this crate owns directly

This facade is intentionally more than a re-export crate. It also owns the integration points that combine multiple domains, including:

the agent pipeline and tool execution loop
provider selection from application configuration
guardrails, checkpoints, compaction, and orchestration helpers
configuration bridges that depend on security or runtime integration
shared surfaces consumed by both CLI and GUI entry points

§High-signal module groups

pipeline: agent request execution, tool routing, reflection, and streaming integration
tools: stable access to built-in tools, schemas, permissions, and streaming shell helpers
config: application configuration plus core-owned security bridges
llm_provider: provider selection and facade access to LLM types
mcp, a2a, knowledge, memory_bank: protocol and knowledge surfaces
session_manager, session_workspace, agent_sessions: session state and workspace lifecycle

§Cargo features

The facade exposes optional capabilities through Cargo features:

voice-local: local Whisper speech-to-text support
nats: NATS messaging integration
json-ld: JSON-LD processing for MDH workflows
security: encryption, keychain, and secure-storage integrations
macos-permissions, linux-permissions, windows-permissions: platform-specific permission helpers

§Documentation strategy

The long-term goal is for crate-level and module-level Rustdoc to become the canonical architecture and API reference surfaced by cargo doc. External documents should increasingly focus on operational workflows such as install, packaging, release, and troubleshooting.

Re-exports§

pub use checkpoints::Checkpoint;
pub use checkpoints::CheckpointError;
pub use checkpoints::CheckpointId;
pub use checkpoints::CheckpointManager;
pub use checkpoints::CheckpointMetadata;
pub use checkpoints::CheckpointRetentionPolicy;
pub use checkpoints::CheckpointSnapshot;
pub use checkpoints::FileCheckpointStore;
pub use checkpoints::default_checkpoints_dir;
pub use compaction::CompactionConfig;
pub use compaction::CompactionEvent;
pub use compaction::CompactionEventType;
pub use compaction::CompactionResult;
pub use compaction::CompactionStrategy;
pub use compaction::ContextCompactor;
pub use config::AppConfigSecurityExt;
pub use llm_provider::select_provider;
pub use pipeline::AgentPipeline;
pub use pipeline::PipelineConfigExt;
pub use speech::SpeechProcessorCoreExt;
pub use speech::get_speech_processor;
pub use speech::is_speech_recording;
pub use speech::update_speech_config;
pub use tools::render_capabilities;
pub use orchestrator::AgentOrchestrator;
pub use orchestrator::OrchestratorAgentManager;
pub use orchestrator::OrchestratorObserver;

Modules§

a2a: Agent-to-Agent protocol types and helpers re-exported from gestura-core-a2a.
agent_sessions: Agent session model + persistence.
agents: Agent lifecycle management and orchestration primitives re-exported from gestura-core-agents.
analytics: Usage analytics types re-exported from gestura-core-analytics.
audio: Audio noise-cancellation utilities re-exported from gestura-core-audio.
audio_capture: Audio capture helpers re-exported from gestura-core-audio.
checkpoints: Checkpoints (snapshots) for agent sessions and related state.
compaction: Context compaction for managing conversation history within token limits
config: Configuration management for Gestura.
config_env: Environment-based configuration loading re-exported from gestura-core-config.
context: Smart request analysis, entity extraction, and context resolution re-exported from gestura-core-context.
default_models: Built-in provider model defaults re-exported from gestura-core-llm.
error: Shared application errors and result aliases re-exported from gestura-core-foundation.
events: Cross-cutting event types re-exported from gestura-core-foundation.
execution_mode: Execution-mode policy primitives re-exported from gestura-core-foundation.
explorer: File-system exploration helpers re-exported from gestura-core-explorer.
gdpr: GDPR-specific helpers re-exported from gestura-core-security.
hooks: Hook engine types re-exported from gestura-core-hooks.
hotkey_ipc: Hotkey inter-process communication types re-exported from gestura-core-ipc.
intent: Unified intent normalization layer re-exported from gestura-core-intent.
interaction: Shared interaction model types re-exported from gestura-core-foundation.
knowledge: Built-in knowledge and expertise surfaces re-exported from gestura-core-knowledge.
llm_overrides: Session-scoped LLM override resolution helpers.
llm_provider: LLM provider selection and integration.
llm_validation: LLM provider/model compatibility helpers.
mcp: Model Context Protocol types and services re-exported from gestura-core-mcp.
memory_bank: Persistent memory-bank types and services re-exported from gestura-core-memory-bank.
memory_console: Shared memory console service for CLI and GUI inspection workflows.
model_display: Human-friendly model display helpers re-exported from gestura-core-foundation.
model_listing: Model discovery and listing helpers re-exported from gestura-core-llm.
nats_mq: NATS messaging primitives re-exported from gestura-core-nats.
openai_compat: OpenAI(-compatible) API compatibility helpers (e.g., parameter support quirks). OpenAI API compatibility helpers.
orchestrator: Subagent orchestration (core-owned, tauri-free).
outcomes: Outcome-linked learning signals re-exported from gestura-core-foundation.
pipeline: Agent Pipeline - Unified LLM interaction pipeline
platform: Platform-detection helpers re-exported from gestura-core-foundation.
plugin_system: Plugin-system types re-exported from gestura-core-plugins.
prompt_enhancement: Prompt enhancement functionality for Gestura
recommendations: Recommendation types re-exported from gestura-core-analytics.
retry: Retry strategy helpers re-exported from gestura-core-retry.
sandbox: Sandbox primitives re-exported from gestura-core-security.
scripting: Scripting-engine types re-exported from gestura-core-scripting.
secrets: Secret-provider abstractions surfaced from foundation and security crates.
security: Security domain types and services re-exported from gestura-core-security.
session_manager: Authentication and session-management services re-exported from gestura-core-sessions.
session_workspace: Session-scoped workspace management re-exported from gestura-core-sessions.
speech: Speech processing integration.
stream_cancellation: Streaming cancellation primitives re-exported from gestura-core-streaming.
stream_error: Streaming error types re-exported from gestura-core-foundation.
stream_health: Streaming health state types re-exported from gestura-core-foundation.
stream_reconnect: Streaming reconnection helpers re-exported from gestura-core-foundation.
streaming: Streaming LLM provider support for Gestura.
stt_provider: Speech-to-text provider interfaces re-exported from gestura-core-audio.
subjects: Common NATS subjects used across the app
tasks: Task-management primitives re-exported from gestura-core-tasks.
telemetry: Telemetry and instrumentation types re-exported from gestura-core-foundation.
token_tracker: Token accounting helpers re-exported from gestura-core-llm.
tool_confirmation: Tool-confirmation flows re-exported from gestura-core-tools.
tool_inspection: Tool inspection and review helpers re-exported from gestura-core-tools.
tools: System Tools for Gestura
workflows: Workflow helpers re-exported from gestura-core-tasks.

Structs§

A2AClient: A2A HTTP Client for agent-to-agent communication
A2AError: A2A Error
A2AMessage: A2A Message
A2ARequest: A2A JSON-RPC Request
A2AResponse: A2A JSON-RPC Response
A2AServer: A2A server for handling incoming JSON-RPC requests.
A2ATask: A2A Task
AgentCard: Agent Card for A2A discovery
AgentCardRegistry: In-memory registry for discovered/known agent cards.
AgentContext: Context hints for provider selection (agent, tenant, etc.)
AgentEnvelope: IPC envelope for events exchanged with agents
AgentInfo: Public agent info for status queries
AgentManager: Core agent manager implementation
AgentProfile: Agent profile for authentication and identity propagation.
AgentRequest: A request to be processed by the agent pipeline
AgentResponse: Response from the agent pipeline
AgentSession: A persisted agent session.
AnalyticsConfig: Analytics configuration
AnalyticsInsights: Analytics insights
AppConfig: Application configuration persisted to a YAML file.
Artifact: Artifact produced by a task
AudioCaptureConfig: Audio capture configuration
AudioDeviceInfo: Audio device information
AuthToken: Authentication token
AuthenticationInfo: Authentication info
CacheStats: Cache statistics
CachedTool: Cached tool information from an MCP server
CancellationToken: Cancellation token for streaming requests
ConfirmationRequest: Confirmation request for user approval
ConsentRecord: Consent record
ContextCache: Smart cache for context data
ContextManager: Manager for handling context in a smart, efficient way
ContextManagerStats: Statistics for the context manager
ConversationMessage: A message in a persisted conversation history.
DataAuditEntry: Data audit entry
DelegatedTask: Task that can be delegated to a subagent
Encryptor: AES-256-GCM encryptor for protecting sensitive data
ErrorAnalysis: Error analysis
EventBufferConfig: Configuration for event buffering
EventEmitter: Event emitter with optional buffering for rate limiting
ExplorerEntry: A single entry returned by list_dir.
ExplorerGitPathStatus: Combined staged/unstaged/untracked status for a single path.
ExplorerGitStatusResponse: Response payload for the git-status query.
ExplorerListDirResponse: Response payload for a directory listing.
ExplorerRootResponse: Response payload for the workspace root query.
ExtractedEntity: An entity extracted from the request
FileAgentSessionStore: File-backed session store (one JSON file per session).
FileContext: File context loaded for the request
GdprManager: GDPR compliance manager
HapticFeedback: Haptic feedback pattern for agent responses
HookContext: Runtime context provided to hooks.
HookEngine: Hook engine.
HookExecutionRecord: A record of a hook execution.
HooksSettings: Global hooks settings.
InspectionResult: Result of tool inspection
InteractionContext: Extended agent context with interaction data
InteractionEvent: Unified interaction event that can trigger agent actions
JsonRpcError: JSON-RPC error payload.
JsonRpcRequest: JSON-RPC request structure.
JsonRpcResponse: JSON-RPC response structure.
KeychainStorage: OS keychain integration (when security feature enabled)
KnowledgeItem: A knowledge item represents a specialized area of expertise that can be loaded on-demand to provide context to the agent.
KnowledgeMatch: Result of matching knowledge items to a query
KnowledgeQuery: Knowledge query for finding relevant knowledge items
KnowledgeReference: A reference file that can be loaded on-demand
KnowledgeSettingsManager: Manager for session-scoped knowledge settings
KnowledgeStore: Knowledge store for managing knowledge items
LlmCallResponse: Response from an LLM call including token usage
LlmResponse: Result of LLM processing
LocalMcp: MCP integrator with token storage and validation
McpCacheStats: Cache statistics
McpClient: An active connection to a single MCP server.
McpClientRegistry: Global registry of active MCP client connections, keyed by server name.
McpDiscoveryManager: MCP Tool Discovery Manager
McpJsonFile: Represents an .mcp.json configuration file.
McpRequestContext: Context passed into tool/resource handlers.
McpServer: Transport-agnostic MCP JSON-RPC server.
McpServerConfig: Configuration for MCP server connection
McpServerEntry: Full MCP server entry compatible with .mcp.json.
McpServerInfo: Information about a connected MCP server
McpToken: Token for MCP authentication
MdhResource: Result of local MDH translation suitable for MCP usage
MemoryBankEntry: A single entry in the memory bank representing a saved conversation context
MemoryBankQuery: Filter options for targeted memory-bank retrieval.
MemoryGovernanceRefreshReport: Summary of a governance refresh pass.
MemoryGovernanceSuggestion: Persisted governance suggestion for operator review.
MemorySearchResult: Ranked result from targeted memory-bank retrieval.
Message: A message in conversation history
Metric: Telemetry metric
MockSecureStorage: Mock secure storage for testing
ModeConfig: Configuration for execution mode behavior
ModeManager: Manager for execution mode state
ModelInfo: A single model entry returned by listing endpoints.
NatsHealthMonitor: NATS connection health monitor
NoiseCancellationConfig: Noise cancellation configuration
NoiseCancellationProcessor: Noise cancellation processor
NoiseReductionStats: Noise reduction statistics
OAuth2Config: OAuth2 configuration
OrchestratorToolCall: Record of a tool call during orchestrated task execution
PausedExecutionState: Captured execution state when an agent session is paused.
PerformanceMetrics: Performance metrics
Permission: A permission grant
PermissionAuditEntry: Audit log entry for a permission check.
PermissionCheck: Permission check result
PermissionManager: Permission management service
PersonalizedRecommendationEngine: Personalized recommendation engine
PipelineConfig: Configuration for the pipeline
PopularMcpServer: A recommended MCP server entry sourced from the official MCP Registry.
ProfileStore: Profile store for managing agent profiles.
ProgressTracker: Progress tracker for long-running operations
Recommendation: Recommendation item
RecommendationConfig: Recommendation engine configuration
ReconnectConfig: Configuration for reconnection behavior
ReconnectManager: Reconnection manager for streaming connections
RegistryBrowseEntry: One entry in a paginated registry browse result.
RegistryBrowsePage: Paginated response returned by browse_mcp_registry.
RequestAnalysis: Request analysis result - determines what context is needed
RequestAnalyzer: Analyzes requests to determine what context is needed
RequestMetadata: Metadata about the request source and context
ResolvedContext: Context that has been resolved and cached
RetryEvent: Retry event for notification callbacks
RetryManager: Retry manager for executing operations with automatic retry
RetryPolicy: Retry policy configuration
SandboxConfig: Sandbox configuration for agent processes
SandboxManager: Sandbox manager for agent processes
Script: Script metadata
ScriptContext: Script execution context
ScriptExecutionResult: Script execution result
ScriptingEngine: Scripting engine
SecureConfigManager: Secure configuration manager with encryption and keychain-backed key storage
SessionInfo: Minimal session info used for list UIs.
SessionKnowledgeSettings: Session-scoped knowledge settings
SessionLlmConfig: Session-scoped LLM configuration override.
SessionManager: Session manager
SessionPatterns: Session usage patterns
SessionReflectionSettings: Session-scoped experiential reflection override.
SessionState: Persisted session state.
SessionToolCall: Tool call record for session history.
SessionToolSettings: Session-scoped tool availability settings.
SessionVoiceConfig: Session-scoped Voice/STT configuration override.
SessionWorkspace: Session workspace configuration
Skill: Skill definition
SpeechConfig: Speech processing configuration
SpeechProcessor: Core speech processor without Tauri dependencies
StreamError: Streaming error with rich context
StreamHealthConfig: Configuration for stream health monitoring
StreamHealthHandle: Lightweight handle for stream health monitoring
StreamHealthMonitor: Stream health monitor
StreamState: Stream state that can be preserved across reconnections
SystemHealth: System health metrics
Task: A task represents a unit of work to be tracked
TaskList: A list of tasks for a session
TaskManager: Task manager for persisting and managing tasks
TaskMemoryEvent: Structured event recorded in task metadata for memory lifecycle tracking.
TaskMemoryLifecycle: Structured task-local memory lifecycle information persisted in metadata.
TaskResult: Result from a delegated task
TelemetryManager: Telemetry manager
TimePeriod: Time period for analysis
Timer: Performance timer for measuring operation duration
TokenInfo: Token information with expiration and permissions
TokenTracker: Token usage tracker for session and global usage
TokenUsage: Token usage information from an LLM API call
ToolCallRecord: Record of a tool call execution
ToolConfirmationInfo: A user-facing description of a tool confirmation prompt.
ToolContext: Minimal tool context for when tools are needed
ToolDefinition: A single tool definition.
ToolHint: Hint for tool selection based on interaction context
ToolInspectionManager: Tool Inspection Manager
ToolMetadata: Tool metadata for inspection
ToolPolicyEvaluation: Evaluation of a tool call.
TranscriptionResult: Result of speech transcription
UnconfiguredProvider: A provider that returns an error when no real provider is configured. Used when config is missing or invalid.
UsageAnalytics: Usage analytics system
UsageEvent: Usage event
UsagePatterns: Usage patterns analysis
UsageRecord: A single token usage record with timestamp
UsageStats: Aggregated token usage statistics
UserBehaviorPattern: User behavior pattern
UserSession: User session information
Workflow: A workflow represents a reusable prompt template
WorkflowInfo: Workflow metadata for listing (without loading full content)
WorkflowManager: Workflow manager for discovering and loading workflow files

Enums§

AgentCommand: Commands that can be sent to an agent task
AgentEvent: Event types for the agent pipeline
AgentStatus: Status value for an agent
AppError: Top-level application error for Gestura.
BudgetStatus: Budget check result
ButtonPressType: Button press types
ConfirmationResponse: User’s response to a confirmation request
ConsentStatus: Consent status for data processing
ContextCategory: Categories of context that might be needed for a request
DataCategory: GDPR data categories
DataOperation: Data operations for audit trail
DispatchEvent: Dispatcher hint for event routing
EntityType: Types of entities that can be extracted
ErrorClass: Error classification for retry decisions
EventType: Usage event types
ExecutionMode: Execution mode for the agent
ExplorerEntryKind: Kind of a directory entry.
ExplorerError: Errors returned by the explorer helpers.
ExplorerGitChangeKind: Kind of change reported by git status --porcelain.
GestureType: Types of gestures that can trigger agent interactions
HapticPattern: Predefined haptic patterns
HookEvent: A hook event.
InteractionType: Types of interactions that can trigger agent actions
KnowledgeError: Error type for knowledge operations
LoadCondition: Conditions for when to load a reference
McpScope: Configuration scope for MCP servers.
McpTransportType: Transport type for MCP server connections.
MemoryBankError: Errors that can occur during memory bank operations
MemoryGovernanceRelationship: Relationship produced by the governance-analysis pass.
MemoryGovernanceState: General governance state for durable memory curation.
MemoryKind: High-level memory retention domain.
MemoryScope: Scope for targeted durable-memory retrieval.
MemoryType: Typed classification for durable memory records.
MessagePart: Message part (text, file, or data)
MessageSource: Source of an end-user message.
MetricType: Metric types for telemetry
PermissionLevel: Permission level for tool execution in a session.
PermissionScope: Scope of a permission
PrivacyMode: Privacy modes for analytics
ProgressStage: Stages of pipeline progress
RecommendationFeedback: User feedback on recommendations
RecommendationType: Recommendation types
ReconnectEvent: Reconnection event for frontend notification
ReconnectState: Reconnection state
ReflectionMemoryState: Retrieval state for reflection-derived durable memories.
RequestSource: Source of the request
ScriptLanguage: Supported scripting languages
ScriptPermission: Script permissions
ScriptTrigger: Script triggers
SecureStorageError: Error type for secure storage operations
ServerState: MCP server connection state
SessionFilter: Filter for listing sessions.
SessionPermissionLevel: Permission level for tool execution.
ShellOutputStream: Which output stream a shell chunk originated from.
ShellProcessState: Lifecycle state of a shell process.
SlideDirection: Direction for slide gestures
StreamChunk: A chunk of streaming response
StreamErrorCategory: Streaming error category for classification
StreamHealthEvent: Stream health event for frontend notification
StreamHealthStatus: Stream health status
TaskError: Error type for task operations
TaskMemoryPhase: Phase in the task/memory lifecycle.
TaskSource: Source of a task (who created it)
TaskStatus: Status of a task
TokenType: Token types
ToolCallDecision: Decision for a tool call at a given permission level.
ToolCategory: Tool category for permission grouping
ToolExecutionCheck: Result of checking if a tool can be executed
ToolPermission: Tool permission level for execution modes
ToolResult: Result of a tool execution
WorkflowError: Error type for workflow operations
WorkspaceError: Errors that can occur during workspace operations

Constants§

DEFAULT_ANTHROPIC_MODEL: Default Anthropic model for agent/completion tasks.
DEFAULT_GEMINI_BASE_URL: Default Gemini API base URL (Google AI Studio / Generative Language API).
DEFAULT_GEMINI_MODEL: Default Gemini model for agent/completion tasks.
DEFAULT_GROK_MODEL: Default Grok model for agent/completion tasks.
DEFAULT_OLLAMA_BASE_URL: Default Ollama base URL for local inference.
DEFAULT_OLLAMA_MODEL: Default Ollama model for local inference.
DEFAULT_OPENAI_MODEL: Default OpenAI model for agent/completion tasks.
DEFAULT_OPENAI_STT_MODEL: Default OpenAI model for speech-to-text transcription.
NAME: Crate name
VERSION: Crate version from Cargo.toml

Traits§

AgentSessionStore: A storage abstraction for agent sessions.
AgentSpawner: Trait for spawning and managing isolated agents
LlmProvider: Unified LLM interface (async)
McpIntegrator: Unifies MCP operations with dual authentication
McpResourceHandler: Handler for an MCP resource.
McpToolHandler: Handler for an MCP tool.
SecureStorage: Secure storage interface for sensitive data

Functions§

all_tools: Return the set of built-in tools.
browse_mcp_registry: Fetch a single page from the MCP Registry with optional full-text search.
canonical_root: Canonicalize the workspace root path.
check_ollama_connectivity: Ping the Ollama endpoint to verify it is reachable.
cleanup_old_sessions: Clean up old session workspaces that are older than the specified duration
clear_memory_bank: Clear all memory bank entries in a workspace
connect_nats: Connect to NATS server
connect_with_retry: Attempt to connect to NATS with retries
create_default_sandbox: Create default sandbox config for different agent types
create_event_channel: Create a new event channel pair
create_gestura_agent_card: Create a default agent card for Gestura.
create_music_noise_canceller: Create a noise cancellation processor with music-optimized settings
create_secure_storage: Create the appropriate secure storage implementation based on features.
create_speech_noise_canceller: Create a noise cancellation processor with speech-optimized settings
default_agent_sessions_dir: Default directory for persisted agent sessions.
detect_system_dark_mode: Detect whether the operating system is configured for dark mode.
ensure_memory_bank_dir: Ensure the memory bank directory exists, creating it if necessary
ensure_safe_rel_path: Validate that rel is a safe, non-escaping relative path.
evaluate_tool_call: Evaluate a tool call against a PermissionLevel.
find_tool: Find a tool definition by name (case-insensitive).
format_anthropic_model_name: Format an Anthropic model ID to a human-readable name.
format_gemini_model_name: Format a Gemini model ID to a human-readable name.
format_grok_model_name: Format a Grok model ID to human-readable name.
format_model_name: Format any model name based on provider.
format_openai_model_name: Format an OpenAI model ID to a human-readable name.
format_token_count: Format token count with K/M suffix
get_gdpr_manager: Get the global GDPR manager
get_global_task_manager: Returns the process-wide shared TaskManager.
get_mcp: Get the global MCP instance
get_mcp_client_registry: Get the global MCP client registry.
get_memory_bank_dir: Get the memory bank directory path for a workspace
get_scripting_engine: Get the global scripting engine
get_session_manager: Get the global session manager
get_sessions_base_dir: Get the base directory for session workspaces
get_telemetry_manager: Get the global telemetry manager
get_token_tracker: Get the global token tracker
import_claude_desktop_servers: Import MCP servers from Claude Desktop config.
init_jetstream: Initialize JetStream context and create KV bucket if missing
is_action_allowed: Return whether an action should be allowed for a session at the given permission level.
is_local_provider: Check if a provider is local (no cost tracking).
is_microphone_available: Check if microphone is available
is_shell_command_write_operation: Conservatively determine whether a shell tool call is likely to perform a write.
is_token_well_formed: Return whether a token is well-formed for Gestura’s current bearer-token scheme.
is_write_operation: Determine if a tool operation is a write operation based on tool name and arguments.
list_audio_input_devices: List all available audio input devices
list_dir: List directory entries under root/dir_rel, returning at most max_entries.
list_memory_bank: List all memory bank entries in a workspace
list_models_for_provider: List available models for the given provider.
list_popular_mcp_servers: Return up to limit popular, open-source MCP servers that can be added without additional configuration.
load_from_memory_bank: Load a memory bank entry from a markdown file
looks_like_capabilities_question: Heuristic: decide whether a user message is asking for full capabilities/config.
looks_like_tools_question: Heuristic: decide whether a user message is asking for a tool inventory.
mdh_translate: Local-only MDH translate: read JSON file, validate minimal JSON-LD shape, and create URI NOTE: In production we will use json-ld-rs to expand/compact; here we simulate locally
normalize_git_change_path: Normalize a git porcelain path into a safe rel_path string.
normalize_mcp_server_name: Convert an arbitrary registry name into a safe, stable tool identifier.
publish_json: Publish a JSON payload to a subject
record_audio: Record audio from the microphone until user stops speaking (4 seconds of silence) Returns the duration of recorded audio in seconds
refresh_memory_bank_governance: Refresh persisted governance suggestions for durable memory entries.
register_builtin_knowledge: Register built-in knowledge items
render_tool_detail: Render a detailed tool description.
render_tools_overview: Render a compact table of built-in tools (static, no config needed).
request_stop_recording: Request the audio recording to stop from external code
requires_confirmation: Return whether an action should require confirmation for a session at the given permission level.
reset_stop_flag: Reset the external stop flag (call before starting a new recording)
resolve_under_root: Resolve rel under root, ensuring the result stays within the root after symlink resolution.
save_to_memory_bank: Save a memory bank entry to disk as a markdown file
search_memory_bank: Search memory bank entries for relevant content
search_memory_bank_with_query: Search memory-bank entries using structured filters and ranking.
spawn_nats_server: Spawn an embedded NATS server with JetStream enabled
start_streaming: Start a streaming LLM request based on config.
subscribe: Subscribe to a subject with message handler
subscribe_wildcard: Subscribe to a wildcard subject

Type Aliases§

AgentSessionResult: Result type for agent session store operations.
EventReceiver: Shared event channel for broadcasting events to multiple listeners
EventSender
NatsConnection: NATS connection type
Result: Result type alias for operations that may fail with AppError.
RetryCallback: Callback type for retry notifications
StreamResult: Result type for streaming operations
WorkspaceResult: Result type for workspace operations