The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.
Rust agent-orchestrationai-agentsai-assistantdesktopllmlocal-firstmcppersonal-aiprivacyrustsecond-braintauri
docs(orchestrator): trim stale delegation notes from tool docs
Steven Enamakel504df3cc2712
diff --git a/crates/openhuman-core/src/tools/orchestrator_tools.rs b/crates/openhuman-core/src/tools/orchestrator_tools.rsindex e26cfa53de..9739de66b6 100644--- a/crates/openhuman-core/src/tools/orchestrator_tools.rs+++ b/crates/openhuman-core/src/tools/orchestrator_tools.rs@@ -5,21 +5,15 @@//! `[subagents]` allowlist constrain that tool; they do not create additional//! model-facing tools.//!-//! For [`SubagentEntry::Skills`] wildcard expansions we synthesise one+//! For [`SubagentEntry::Skills`] wildcard expansions we synthesize one//! `ToolExposure::Deferred` [`ComposioActionTool`] per action of every//! connected Composio toolkit. Those never reach the wire: a belt that opted//! into discovery (`tool_search` in its `[tools] named`) finds them through-//! the harness's `tool_search` bridge and calls them directly, so "send this-//! email" is one search and one call. There is no `delegate_to_integrations_-//! agent` any more — routing a single action through a sub-agent spawn cost a-//! blocking agentic round-trip and a second prompt for work the parent could-//! do in one call.+//! the harness's `tool_search` bridge and calls them directly.//!//! Called from [`crate::agent::session_host::builder`] at-//! agent-build time, with the orchestrator's own definition, the global-//! definition and the current list of connected Composio integrations.-//!-//! [`AgentDefinition::when_to_use`]: crate::agent_registry::host_definition::AgentDefinition::when_to_use+//! agent-build time, with the orchestrator's own definition and the current+//! list of connected Composio integrations.//! [`SubagentEntry::AgentId`]: tinyagents_definition::SubagentEntry::AgentId//! [`SubagentEntry::Skills`]: tinyagents_definition::SubagentEntry::Skills@@ -30,15 +24,10 @@ use tinyagents_definition::SubagentEntry;use tinytools::Tool;-/// Synthesise the delegation tool list for an agent based on its-/// declarative `subagents` field.+/// Return deferred integration actions for the agent's skills wildcard.///-/// Each [`SubagentEntry::AgentId`] is resolved against `registry` and-/// rendered as an [`ArchetypeDelegationTool`] whose `name()` defaults to-/// `delegate_{target.id}` (overridable via the target agent's-/// `delegate_name` field) and whose `description()` is the target's-/// `when_to_use` — so editing an agent's TOML description immediately-/// updates the tool schema the orchestrator LLM sees, with zero drift.+/// `AgentId` entries only configure the `spawn_subagent` allowlist; they do+/// not create additional model-facing tools.////// Each [`SubagentEntry::Skills`] wildcard expands to the connected/// integrations' actions as `Deferred` tools@@ -52,10 +41,7 @@ use tinytools::Tool;/// `connected_integrations` slice produces zero tools, which is the correct/// behaviour when the user has not yet connected any integrations.///-/// Returns an empty Vec when `definition.subagents` is empty — callers-/// (notably the builder) handle this by not extending the visible-tool-/// set, so non-delegating agents behave identically to how they did-/// before this module existed.+/// Returns an empty Vec when there are no connected integration actions.pub fn collect_orchestrator_tools(definition: &AgentDefinition,connected_integrations: &[ConnectedIntegration],@@ -86,10 +72,8 @@ pub fn collect_orchestrator_tools(// never sees them — the session builder leaves them// prompt-hidden, which the direct-call gate refuses. Approval// and channel permission apply per call. This used to sit- // beside a collapsed `delegate_to_integrations_agent` tool- // that spawned `integrations_agent` per toolkit; with the- // actions searchable that spawn only added a blocking- // sub-agent round-trip, so the delegation tool is gone.+ // beside a per-toolkit sub-agent route; searchable actions+ // are directly callable through the harness.let actions = collect_deferred_integration_actions(connected_integrations);if !actions.is_empty() {log::debug!(