feat(terminals): reprise de conversation par cellule + fix ordre d'écriture

Permet de recharger la conversation CLI précédente de chaque cellule à la
réouverture du projet, de façon universelle (indépendant du modèle/CLI).

- profil AgentRuntime: bloc déclaratif optionnel `session { assignFlag, resumeFlag }`
- LeafCell: `conversationId` (persistant, distinct du SessionId PTY) + `agentWasRunning`
- runtime: SessionPlan (None/Assign/Resume) + composition pure des args
- LaunchAgent: décide Assign vs Resume, génère l'UUID, remonte l'id assigné
  (persistance par l'appelant via setCellConversation — découplage SRP)
- close: SnapshotRunningAgents fige `agentWasRunning` avant le kill-all
  (statut clot/en cours universel, sans parsing CLI)
- SessionInspector: port optionnel best-effort + adapter ClaudeTranscriptInspector
- popup de reprise par cellule (statut + sujet/tokens si dispo), intercalée
  avant le Resume auto, jamais sur le chemin reattach

fix(terminals): sérialise les écritures PTY (file FIFO par handle) — corrige
les caractères mélangés/accents dus au réordonnancement des invoke Tauri concurrents

fix(layout): l'opération `move` préservait mal les champs du leaf (perdait `agent`)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

crates/app-tauri/src/commands.rs

@@ -11,9 +11,10 @@ use application::{
     AppError, CloseProjectInput, CreateAgentInput, CreateLayoutInput, DeleteAgentInput,
     DeleteLayoutInput, DeleteTemplateInput, DetectAgentDriftInput, GitBranchesInput,
     GitCheckoutInput, GitCommitInput, GitGraphInput, GitInitInput, GitLogInput, GitStagePathInput,
-    GitStatusInput, LaunchAgentInput, ListAgentsInput, ListLayoutsInput, LoadLayoutInput,
+    GitStatusInput, InspectConversationInput, LaunchAgentInput, ListAgentsInput, ListLayoutsInput, LoadLayoutInput,
     MutateLayoutInput, OpenProjectInput, ReadAgentContextInput, RenameLayoutInput,
-    SetActiveLayoutInput, SyncAgentWithTemplateInput, UpdateAgentContextInput,
+    SetActiveLayoutInput, SnapshotRunningAgentsInput, SyncAgentWithTemplateInput,
+    UpdateAgentContextInput,
     AssignSkillToAgentInput, CreateSkillInput, DeleteSkillInput, ListSkillsInput,
     UnassignSkillFromAgentInput, UpdateSkillInput,
 };
@@ -28,6 +29,7 @@ use crate::dto::{
     DetectProfilesRequestDto, DetectProfilesResponseDto, ErrorDto, FirstRunStateDto,
     GitBranchesDto, GitCheckoutRequestDto, GitCommitDto, GitCommitListDto, GitCommitRequestDto,
     GitStageRequestDto, GitStatusListDto, GraphCommitListDto, HealthRequestDto, HealthResponseDto,
+    ConversationDetailsDto, InspectConversationRequestDto,
     LaunchAgentRequestDto, LayoutDto, LayoutOperationDto, ListLayoutsDto, OpenTerminalRequestDto,
     ProfileDto, ProfileListDto, ProjectDto, ProjectListDto, ReadAgentContextResponseDto,
     ReattachResultDto, RenameLayoutRequestDto, ResizeTerminalRequestDto, SaveProfileRequestDto,
@@ -110,6 +112,13 @@ pub async fn close_project(
     state: State<'_, AppState>,
 ) -> Result<(), ErrorDto> {
     let id = parse_project_id(&project_id)?;
+    // T5: freeze `agent_was_running` on every agent leaf BEFORE any PTY release,
+    // reading the live registry as it stands now. Best-effort: a snapshot failure
+    // must not block the close.
+    let _ = state
+        .snapshot_running_agents
+        .execute(SnapshotRunningAgentsInput { project_id: id })
+        .await;
     // Stop tailing this project's `.ideai/requests/` tree (§14.3).
     state.stop_orchestrator_watch(&id);
     state
@@ -740,6 +749,35 @@ pub async fn delete_agent(
         .map_err(ErrorDto::from)
 }
 
+/// `inspect_conversation` — best-effort enriched details (last topic + token
+/// indicator) for a resume popup (T7).
+///
+/// Best-effort by contract: a missing/unsupported inspector or a missing
+/// transcript yields **empty details** (absent `lastTopic`/`tokenCount`), never
+/// an error. Only a genuine store failure (resolving the agent/profile) errors.
+///
+/// # Errors
+/// Returns an [`ErrorDto`] (`INVALID` for a malformed id, `NOT_FOUND` if the
+/// project, agent or profile is unknown, `STORE` on a manifest/profile failure).
+#[tauri::command]
+pub async fn inspect_conversation(
+    request: InspectConversationRequestDto,
+    state: State<'_, AppState>,
+) -> Result<ConversationDetailsDto, ErrorDto> {
+    let project = resolve_project(&request.project_id, &state).await?;
+    let agent_id = parse_agent_id(&request.agent_id)?;
+    state
+        .inspect_conversation
+        .execute(InspectConversationInput {
+            project,
+            agent_id,
+            conversation_id: request.conversation_id,
+        })
+        .await
+        .map(ConversationDetailsDto::from)
+        .map_err(ErrorDto::from)
+}
+
 /// `launch_agent` — spawn an agent's CLI in a PTY and wire its byte stream to
 /// the frontend via a [`Channel`].
 ///
@@ -770,6 +808,10 @@ pub async fn launch_agent(
             rows: request.rows,
             cols: request.cols,
             node_id: None,
+            // Resume id is a property of the hosting cell; the frontend passes the
+            // leaf's current conversation id here. `None` until the launch_agent
+            // command is made layout-aware (T4 wiring follow-up).
+            conversation_id: request.conversation_id.clone(),
         })
         .await
         .map_err(ErrorDto::from)?;