bamboo_engine/session_app/child_session/

mod.rs

//! Child session management use cases.
//!
//! Provides the application-layer logic for managing child sessions within
//! a root session. The server layer implements `ChildSessionPort` to supply
//! the infrastructure operations (load, save, schedule, cancel).

use async_trait::async_trait;
use bamboo_domain::session::runtime_state::ChildWaitPolicy;
use bamboo_domain::Session;
use std::collections::HashMap;

mod actions;
mod helpers;

#[cfg(test)]
mod tests;

pub use actions::{
    assemble_session_tree, build_session_tree_action, cancel_child_action, create_child_action,
    delete_child_action, get_child_action, list_children_action, run_child_action,
    send_message_to_child_action, update_child_action, SessionTreeNode,
};
pub use helpers::{
    compute_status_guidance, format_child_assignment, map_child_entry, metadata_text,
    normalize_non_empty_optional, normalize_required_text, replace_or_append_last_user_message,
    truncate_after_index, truncate_after_last_user,
};

// ---------------------------------------------------------------------------
// Error type
// ---------------------------------------------------------------------------

#[derive(Debug, thiserror::Error)]
pub enum ChildSessionError {
    #[error("session not found: {0}")]
    NotFound(String),
    #[error("session is not a root session: {0}")]
    NotRootSession(String),
    #[error("session is not a child session: {0}")]
    NotChildSession(String),
    #[error("child session {child_id} does not belong to parent {parent_id}")]
    NotChildOfParent { child_id: String, parent_id: String },
    #[error("{0}")]
    InvalidArguments(String),
    #[error("{0}")]
    Execution(String),
}

// ---------------------------------------------------------------------------
// Value types
// ---------------------------------------------------------------------------

/// Summary of a child session for listing.
#[derive(Debug, Clone)]
pub struct ChildSessionEntry {
    pub child_session_id: String,
    pub title: String,
    pub pinned: bool,
    pub message_count: usize,
    pub updated_at: String,
    pub last_run_status: Option<String>,
    pub last_run_error: Option<String>,
}

/// Result of deleting a child session.
#[derive(Debug, Clone)]
pub struct DeleteChildResult {
    pub deleted: bool,
    pub cancelled_running_child: bool,
}

/// Diagnostic snapshot of a running child session runner.
#[derive(Debug, Clone)]
pub struct ChildRunnerInfo {
    pub started_at: Option<chrono::DateTime<chrono::Utc>>,
    pub completed_at: Option<chrono::DateTime<chrono::Utc>>,
    pub last_tool_name: Option<String>,
    pub last_tool_phase: Option<String>,
    pub last_event_at: Option<chrono::DateTime<chrono::Utc>>,
    pub round_count: u32,
}

/// Result of a logical parent→child delivery.
///
/// Both variants mean the envelope is already durable. `ActivationPending`
/// preserves the enqueue-success/wake-failure distinction so a tool caller can
/// report success and let restart recovery retry the durable activation instead
/// of generating a second message id.
#[derive(Debug)]
pub enum ChildSessionMessageDelivery {
    Activated(crate::SessionMessengerReceipt),
    ActivationPending {
        delivery: bamboo_domain::SessionInboxReceipt,
        error: String,
    },
}

/// Short delegation note appended to the full root-style base prompt for a
/// sub-agent. A sub-agent is a first-class agent (same base prompt + context
/// enhancement as a top-level session); this note only frames the delegation
/// relationship and the expected concise hand-back to the parent.
pub const DELEGATION_NOTE: &str = r#"---

You are running as a delegated sub-agent, spawned by a parent agent to handle a focused task. You have the full capabilities of a top-level agent, including the ability to spawn your own sub-agents when that helps. Stay focused on the assigned task, and when you finish, report a concise conclusion first (the parent receives your final message as the task result)."#;

/// Input for creating a child session.
#[derive(Debug, Clone)]
pub struct CreateChildInput {
    pub parent_session: Session,
    pub child_id: String,
    pub title: String,
    pub responsibility: String,
    pub assignment_prompt: String,
    pub subagent_type: String,
    /// Absolute path to the working directory for the child session.
    pub workspace: String,
    /// How the child workspace was selected before validation.
    pub workspace_source: crate::project_context::WorkspaceSource,
    /// Optional model override resolved from subagent_type routing.
    /// When `None`, the child inherits the parent session's model.
    pub model_override: Option<String>,
    /// Optional provider+model override resolved from subagent routing.
    /// When present, this preserves cross-provider routing for child execution.
    pub model_ref_override: Option<bamboo_domain::ProviderModelRef>,
    /// Runtime metadata resolved from subagent routing (e.g. external agent config).
    pub runtime_metadata: std::collections::HashMap<String, String>,
    /// Whether to immediately enqueue the child for execution.
    /// Defaults to `true`.
    pub auto_run: bool,
    /// Optional reasoning effort to apply to the child's own LLM calls.
    /// `None` (the default) leaves `Session::reasoning_effort` at `None`,
    /// so the provider falls back to its default. The child does NOT
    /// inherit the parent's reasoning_effort — fan-out children that
    /// only need a quick lookup should not pay for `xhigh` reasoning
    /// just because the orchestrator is running at `xhigh`.
    pub reasoning_effort: Option<bamboo_domain::ReasoningEffort>,
    /// Lifecycle of this child: `Some("resident")` marks a reusable resident
    /// agent (one stable session reused for successive tasks under the same
    /// root); `None`/`Some("oneshot")` is the default throwaway child.
    pub lifecycle: Option<String>,
    /// For a resident agent, the stable reuse key (scoped to the root session).
    pub resident_name: Option<String>,
    /// For a resident agent, how successive tasks treat prior context:
    /// `"reset"` (default — independent tasks) or `"accumulate"` (remember).
    pub resident_context: Option<String>,
    /// Tool names to disable for this child (denylist; matched by EXACT
    /// `ToolSchema.function.name`). `None` (the default) = full toolset. A
    /// read-only Guardian reviewer sets e.g. {"Edit","Write","SubAgent",...}.
    /// Carried to the child's `SpawnJob.disabled_tools` via the child session
    /// metadata (see `create_child_action`) so the worker trims its toolset.
    pub disabled_tools: Option<std::collections::BTreeSet<String>>,
    /// Model-controllable context fork (Phase 3): when `Some(n)` with `n > 0`,
    /// the last `n` non-system parent messages are rendered into a "Forked
    /// context from parent" block prepended to the child's task brief. `None`
    /// (the default) keeps the child on a clean, freshly-seeded context.
    pub context_fork: Option<usize>,
}

/// Result of creating a child session.
#[derive(Debug, Clone)]
pub struct CreateChildResult {
    pub child_session_id: String,
    pub model: String,
}

/// A queued follow-up message stored in session metadata for later injection.
#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
pub struct QueuedInjectedMessage {
    pub content: String,
    #[serde(default)]
    pub created_at: Option<chrono::DateTime<chrono::Utc>>,
}

// ---------------------------------------------------------------------------
// Port trait
// ---------------------------------------------------------------------------

#[async_trait]
pub trait ChildSessionPort: Send + Sync {
    /// Validate and normalize the child's workspace before any child/session
    /// state is created. Server adapters override this with the authoritative
    /// Project registry ownership check; non-server embeddings still apply the
    /// shared confinement resolver.
    async fn validate_child_workspace(
        &self,
        _project_id: Option<&bamboo_domain::ProjectId>,
        requested_workspace: &str,
    ) -> Result<String, ChildSessionError> {
        normalize_child_workspace(requested_workspace)
    }

    /// Publish a child workspace that already passed this port's validation.
    ///
    /// Server adapters override this so validation and publication use the
    /// same AppState-scoped confinement resolver. The default preserves the
    /// process-global behavior for non-server embeddings.
    fn publish_child_workspace(
        &self,
        session_id: &str,
        workspace: std::path::PathBuf,
        source: &str,
    ) -> std::path::PathBuf {
        let _ = source;
        bamboo_agent_core::workspace_state::publish_resolved_workspace(session_id, workspace)
    }

    async fn load_root_session(&self, root_id: &str) -> Result<Session, ChildSessionError>;
    async fn load_child_for_parent(
        &self,
        parent_id: &str,
        child_id: &str,
    ) -> Result<Session, ChildSessionError>;
    async fn save_child_session(&self, child: &mut Session) -> Result<(), ChildSessionError>;
    /// Save a child session whose `agent_runtime_state` posture
    /// (`permission_mode` / `no_human_approver`) the caller just set
    /// authoritatively (the #74 resident-reuse re-seed) — persists them as-is
    /// instead of adopting the child's stale on-disk value, unlike
    /// [`Self::save_child_session`], which protects a concurrent `PATCH` to a
    /// running child. Use ONLY right after deliberately writing those flags. #540.
    async fn save_child_session_authoritative_flags(
        &self,
        child: &mut Session,
    ) -> Result<(), ChildSessionError>;
    /// Deliver one parent→child peer message through the runtime-owned logical
    /// SessionMessenger. Implementations must not mutate the Session snapshot
    /// to enqueue.
    async fn send_session_message(
        &self,
        source_session_id: &str,
        target_session_id: &str,
        message: &str,
        idempotency_key: Option<&str>,
    ) -> Result<ChildSessionMessageDelivery, ChildSessionError> {
        let _ = (
            source_session_id,
            target_session_id,
            message,
            idempotency_key,
        );
        Err(ChildSessionError::Execution(
            "logical SessionMessenger is not configured for this runtime".to_string(),
        ))
    }
    /// Commit the live parent's posture plus a validated workspace when
    /// reusing a resident. Persistence happens before the runtime workspace is
    /// published, so a failed save cannot move tools onto an uncommitted path.
    async fn save_resident_reuse_state(
        &self,
        child: &mut Session,
        workspace: &str,
        workspace_source: crate::project_context::WorkspaceSource,
        permission_audit: bamboo_domain::PermissionAuditSeed,
        no_human_approver: bool,
    ) -> Result<(), ChildSessionError> {
        let previous_mode = child
            .agent_runtime_state
            .as_ref()
            .map(|state| state.effective_permission_mode())
            .unwrap_or_default();
        let previous_resolution =
            bamboo_domain::PermissionAuditSnapshot::from_metadata(&child.metadata)
                .map(|snapshot| snapshot.resolution);
        child.workspace = Some(workspace.to_string());
        child.set_workspace_path_meta(workspace);
        child.metadata.insert(
            crate::project_context::WORKSPACE_SOURCE_METADATA_KEY.to_string(),
            workspace_source.as_str().to_string(),
        );
        let runtime = child
            .agent_runtime_state
            .get_or_insert_with(bamboo_domain::AgentRuntimeState::default);
        runtime.set_permission_mode(permission_audit.resolution.requested);
        runtime.no_human_approver = no_human_approver;
        let changed = previous_mode != permission_audit.resolution.requested;
        let posture_changed = previous_resolution != Some(permission_audit.resolution);
        let transitioned_at = posture_changed.then(|| chrono::Utc::now().to_rfc3339());
        bamboo_domain::record_permission_audit(
            &mut child.metadata,
            &permission_audit,
            transitioned_at.as_deref(),
        )
        .map_err(|error| ChildSessionError::Execution(error.to_string()))?;
        if changed {
            child.metadata_version = child.metadata_version.saturating_add(1);
        }
        self.save_child_session_authoritative_flags(child).await?;
        self.publish_child_workspace(
            &child.id,
            std::path::PathBuf::from(workspace),
            workspace_source.as_str(),
        );
        Ok(())
    }
    async fn is_child_running(&self, child_id: &str) -> bool;
    async fn list_children(&self, parent_id: &str) -> Vec<ChildSessionEntry>;
    async fn enqueue_child_run(
        &self,
        parent: &Session,
        child: &Session,
    ) -> Result<(), ChildSessionError>;
    async fn cancel_child_run_and_wait(&self, child_id: &str) -> Result<(), ChildSessionError>;
    async fn delete_child_session(
        &self,
        parent_id: &str,
        child_id: &str,
    ) -> Result<DeleteChildResult, ChildSessionError>;
    /// Return live diagnostic info for a running child session, if available.
    async fn get_child_runner_info(&self, child_id: &str) -> Option<ChildRunnerInfo>;

    /// Register a durable parent wait for a single enqueued child. Idempotent
    /// and coalesced per parent (concurrent sibling spawns merge into one write).
    async fn register_parent_wait_for_child(
        &self,
        parent_session_id: &str,
        child_session_id: &str,
        tool_call_id: Option<&str>,
    ) -> Result<(), ChildSessionError>;

    /// Register a durable parent wait over an explicit set of children with a
    /// chosen policy (the `SubAgent.wait` action). Returns the number of
    /// children the wait now covers (0 = nothing to wait on).
    async fn register_parent_wait_for_children(
        &self,
        parent_session_id: &str,
        child_session_ids: &[String],
        policy: ChildWaitPolicy,
    ) -> Result<usize, ChildSessionError>;

    /// The parent's currently-active (non-terminal) child session ids.
    async fn active_child_ids(&self, parent_session_id: &str) -> Vec<String>;

    /// The subset of `candidates` the session index POSITIVELY reports as
    /// terminal children of this parent, as `(child_id, status)` pairs
    /// (issue #546). `SubAgent.wait` uses this to avoid arming a wait over a
    /// child that fires no further completion — and, policy-permitting, to
    /// short-circuit a wait the terminal statuses already satisfy. Unknown
    /// ids are NOT reported (index-less backends can't distinguish "terminal"
    /// from "not yet indexed"); the child-wait watchdog rescues those at
    /// runtime. Default: empty (no filtering).
    async fn terminal_child_ids(
        &self,
        parent_session_id: &str,
        candidates: &[String],
    ) -> Vec<(String, String)> {
        let _ = (parent_session_id, candidates);
        Vec::new()
    }

    /// Find an existing resident agent in the same root tree by its stable
    /// `resident_name`, returning its child session id if one exists. Used to
    /// reuse a resident agent for a new task instead of minting a new child.
    /// Index-backed (matches `root_session_id` + `metadata["resident_name"]`).
    async fn find_resident_child(
        &self,
        root_session_id: &str,
        resident_name: &str,
    ) -> Option<String>;

    /// Best-effort: ensure the child's session-index entry is visible
    /// immediately after creation (the index is otherwise eventually
    /// consistent). Failures are ignored by the caller.
    async fn ensure_child_indexed(&self, child_session_id: &str);
}

fn normalize_child_workspace(requested_workspace: &str) -> Result<String, ChildSessionError> {
    let requested_workspace = requested_workspace.trim();
    if requested_workspace.is_empty() {
        return Err(ChildSessionError::InvalidArguments(
            "child workspace must be a non-empty path".to_string(),
        ));
    }
    let requested = std::path::PathBuf::from(requested_workspace);
    if requested.exists() && !requested.is_dir() {
        return Err(ChildSessionError::InvalidArguments(format!(
            "child workspace is not a directory: {requested_workspace}"
        )));
    }
    let canonical = requested.canonicalize().unwrap_or(requested);
    let final_workspace = bamboo_agent_core::workspace_state::resolve_workspace_path(canonical);
    Ok(bamboo_config::paths::path_to_display_string(
        &final_workspace,
    ))
}

// ---------------------------------------------------------------------------
// Subagent resolution port
// ---------------------------------------------------------------------------

/// Resolves subagent-type–specific configuration (model, runtime metadata)
/// for the `SubAgent` tool.
///
/// Kept separate from [`ChildSessionPort`] (session CRUD/lifecycle/state): this
/// port is pure `subagent_type` → config resolution (cross-provider model
/// routing + actor/external-agent metadata). The server layer implements it;
/// the tool depends only on the trait, carrying no `AppState` coupling.
#[async_trait]
pub trait SubagentResolutionPort: Send + Sync {
    /// Provider+model ref for a `subagent_type`, or `None` to use defaults.
    async fn resolve_subagent_model(
        &self,
        subagent_type: &str,
    ) -> Option<bamboo_domain::ProviderModelRef>;

    /// Runtime metadata (e.g. external-agent routing) for a `subagent_type`.
    async fn resolve_runtime_metadata(&self, subagent_type: &str) -> HashMap<String, String>;
}

/// Models available from one configured provider (best-effort listing).
#[derive(Debug, Clone, serde::Serialize)]
pub struct ProviderModelList {
    pub provider: String,
    pub models: Vec<String>,
    /// Set when this provider's listing failed (auth missing, network, …);
    /// the provider is still usable with an explicitly known model id.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub error: Option<String>,
}

/// Lists the models the parent can pin a child session to
/// (`SubAgent` tool `action=list_models` / `create.model`).
///
/// Separate from [`SubagentResolutionPort`] because it is backed by the live
/// provider registry rather than per-`subagent_type` config resolution.
#[async_trait]
pub trait ModelCatalogPort: Send + Sync {
    /// Best-effort model listing per configured provider. Providers whose
    /// listing fails are still returned (with `error` set) so the caller can
    /// see they exist.
    async fn list_models(&self) -> Vec<ProviderModelList>;

    /// The default provider name (used to resolve a bare model id without a
    /// `provider:` prefix).
    fn default_provider(&self) -> String;
}