Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion crates/gaviero-tui/CLAUDE.md
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ Workspace dispatch: directory → `Workspace::single_folder`; `*.gaviero-workspa

**Observer bridge:** implements `WriteGateObserver` and `AcpObserver` from [`gaviero_core::observer`](../gaviero-core/src/observer.rs), plus the memory / manifest / MCP observers. Each holds an event-channel sender. **No background task mutates `App` directly.**

**SWARM panel** (`Alt+w`): [`panels/swarm_dashboard.rs`](src/panels/swarm_dashboard.rs) holds state and rendering, [`app/swarm_run.rs`](src/app/swarm_run.rs) launching and input. Three views: script picker (`ScriptCatalog`, plus a path typed in the filter), input form (`gaviero_dsl::inspect_file`, plus a Profile selector when any model profile exists; its Prompt field is the chat's box, [`widgets/prompt_input.rs`](src/widgets/prompt_input.rs), and `$skill` is expanded at launch; `file` and `dir` vars complete like `@`, and list params complete the tier after `id=` from the selected profile; below the vars, when the resolved profile reads, an **Agents** section (one `FieldRole::AgentTier` choice per agent the workflow runs — its tier, from the profile's tiers → `--tier`) and a **Tiers** section (per tier those agents now use, a `FieldRole::TierModel` choice — its client, from the profile's clients → `--tier-client` — and a `FieldRole::TierEffort` choice — the chosen model's levels from core `effort_completions` (`SwarmForm::effort_options`, stubbed in tests) plus the client's own effort or `(provider default)` → `--tier-effort`); on these rows Enter opens the options as a list in the form's popup (`AutocompleteMode::Choice`, `open_choice_list`), ↑/↓ move, Enter picks and moves to the next field, ←/→/Space still cycle; only changed rows are passed, the profile is re-read (`reload_profile_info`) when Folder or Profile changes, and choices survive where the new profile still offers them; the form opens on its first input, below the Folder / Workflow / Profile selectors; a list param is `FieldValue::List`, one input per item in `--param` entry syntax, split / rejoined by `split_list_value` / `join_list_items` — short form unless an item is a `{…}` literal, so free fields survive), and the run (one row per planned agent; the selected agent's output below). A run is `pipeline::execute` in-process, prepared by the same [`script_run`](../gaviero-core/src/swarm/script_run.rs) helpers as `gaviero-cli --script`; keep the two in step. It uses the pipeline's Write Gate, not TURN REVIEW. Every update carries a run id so a stopped run's stragglers are dropped. This panel is the only way the TUI runs a script: chat has no `/run`, and `/swarm`, `/cswarm` and `/undo-swarm` stay gone. Remote does not mirror the panel.
**SWARM panel** (`Alt+w`): [`panels/swarm_dashboard.rs`](src/panels/swarm_dashboard.rs) holds state and rendering, [`app/swarm_run.rs`](src/app/swarm_run.rs) launching and input. Three views: script picker (`ScriptCatalog`, plus a path typed in the filter), input form (`gaviero_dsl::inspect_file`, plus a Profile selector when any model profile exists; its Prompt field is the chat's box, [`widgets/prompt_input.rs`](src/widgets/prompt_input.rs), and `$skill` is expanded at launch; `file` and `dir` vars complete like `@`, and list params complete the tier after `id=` from the selected profile; below the vars, when the resolved profile reads, an **Agents** section (one `FieldRole::AgentTier` choice per agent the workflow runs — its tier, from the profile's tiers → `--tier`) and a **Tiers** section (per tier those agents now use, a `FieldRole::TierModel` choice — its client, from the profile's clients → `--tier-client` — and a `FieldRole::TierEffort` choice — the chosen model's levels from core `effort_completions` (`SwarmForm::effort_options`, stubbed in tests) plus the client's own effort or `(provider default)` → `--tier-effort`); on these rows Enter opens the options as a list in the form's popup (`AutocompleteMode::Choice`, `open_choice_list`), ↑/↓ move, Enter picks and moves to the next field, ←/→/Space still cycle; only changed rows are passed, the profile is re-read (`reload_profile_info`) when Folder or Profile changes, and choices survive where the new profile still offers them; the form opens on its first input, below the Folder / Workflow / Profile selectors; a list param is `FieldValue::List`, one input per item in `--param` entry syntax, split / rejoined by `split_list_value` / `join_list_items` — short form unless an item is a `{…}` literal, so free fields survive), and the run (one row per planned agent; the selected agent's output below). Starting a run moves the form into `SwarmDashboardState::last_forms` (keyed by script path, this session only): opening that script again — `r` on a finished or stopped run, or the picker — restores every value (`SwarmForm::restore_from`: selectors by option name, agent/tier/effort rows, typed inputs the script still has) with focus on Run. A run is `pipeline::execute` in-process, prepared by the same [`script_run`](../gaviero-core/src/swarm/script_run.rs) helpers as `gaviero-cli --script`; keep the two in step. It uses the pipeline's Write Gate, not TURN REVIEW. Every update carries a run id so a stopped run's stragglers are dropped. This panel is the only way the TUI runs a script: chat has no `/run`, and `/swarm`, `/cswarm` and `/undo-swarm` stay gone. Remote does not mirror the panel.

**Turn review (mandatory):** [`app/turn_review.rs`](src/app/turn_review.rs) + the dispatch task in [`app/side_panel.rs`](src/app/side_panel.rs). Each chat turn is bracketed by `app.turn_capture.begin/end` (core `turn_capture`); the session runs with `AgentOptions::host_capture` and the conversation's gate mode `AutoAccept`. A non-empty change set opens the TURN REVIEW left panel (non-modal) and blocks *that* conversation's next prompt (desktop keeps the typed prompt; remote gets `turn_review_pending`). Pending reviews and their decisions persist under `.gaviero/turns/pending/` and reopen on startup. Any host write made while a turn may be running — editor saves, review reverts, MCP config synthesis — must go through `turn_review::note_host_write` or it is attributed to the agent.

Expand Down
82 changes: 80 additions & 2 deletions crates/gaviero-tui/src/app/swarm_run.rs
Original file line number Diff line number Diff line change
Expand Up @@ -243,8 +243,12 @@ fn open_form(app: &mut App, script: ScriptRow) {
let profiles =
ProfileCatalog::scan(&app.workspace, &ScriptCatalog::global_scripts_dir())
.names();
app.swarm_dashboard.form =
Some(SwarmForm::new(script, inspection, folders, selected).with_profiles(profiles));
let mut form =
SwarmForm::new(script, inspection, folders, selected).with_profiles(profiles);
if let Some(previous) = app.swarm_dashboard.last_forms.get(&form.script.path) {
form.restore_from(previous);
}
app.swarm_dashboard.form = Some(form);
app.swarm_dashboard.view = SwarmView::Form;
reload_profile_info(app);
refresh_form_warnings(app);
Expand Down Expand Up @@ -1156,6 +1160,17 @@ fn handle_run(app: &mut App, action: Action) {
show_selected_diff(app);
return;
}
Action::InsertChar('r') if !dash.running => {
let script = dash
.last_script
.as_ref()
.and_then(|path| dash.last_forms.get(path))
.map(|form| form.script.clone());
if let Some(script) = script {
open_form(app, script);
}
return;
}
Action::InsertChar('u') if !dash.running => {
undo_pressed(app);
return;
Expand Down Expand Up @@ -1753,6 +1768,69 @@ tier spare other
assert!(notes.contains("Tier local runs on client other"), "{notes}");
}

#[tokio::test(flavor = "multi_thread")]
async fn running_a_script_again_starts_from_the_values_of_its_last_run() {
let dir = tempfile::tempdir().unwrap();
let root = dir.path().canonicalize().unwrap();
let (url, _asked) = fake_ollama("custom_out/report.md");
let (mut app, mut rx) = app_with_script(&root, &url);

open_filled_form(&mut app);
handle_action(&mut app, Action::Tab);
for _ in 0.."default_out".len() {
handle_action(&mut app, Action::Backspace);
}
type_text(&mut app, "custom_out");
press_run(&mut app);
assert_eq!(app.swarm_dashboard.view, SwarmView::Run);
pump_until(&mut app, &mut rx, |a| !a.swarm_dashboard.running).await;

// `r` on the finished run: the same values, ready to run again.
handle_action(&mut app, Action::InsertChar('r'));
assert_eq!(app.swarm_dashboard.view, SwarmView::Form);
assert_eq!(field_text(&app, "Prompt"), "the moon");
assert_eq!(field_text(&app, "OUT_DIR"), "custom_out");
assert!(app.swarm_dashboard.form.as_ref().unwrap().on_run_button());

// Picking the script from the list restores them too.
handle_escape(&mut app);
assert_eq!(app.swarm_dashboard.view, SwarmView::Picker);
handle_action(&mut app, Action::Enter);
assert_eq!(field_text(&app, "Prompt"), "the moon");
assert_eq!(field_text(&app, "OUT_DIR"), "custom_out");
}

#[tokio::test(flavor = "multi_thread")]
async fn a_stopped_run_keeps_its_values_for_the_next_one() {
let dir = tempfile::tempdir().unwrap();
let root = dir.path().canonicalize().unwrap();
let (mut app, mut rx) = app_with_script(&root, &silent_ollama());

open_filled_form(&mut app);
press_run(&mut app);
pump_until(&mut app, &mut rx, |a| {
matches!(a.swarm_dashboard.agents[0].status, AgentStatus::Running)
})
.await;
assert!(
handle_action_is_ignored_while_running(&mut app),
"`r` does nothing while the run is going"
);
handle_action(&mut app, Action::InsertChar('x'));
handle_action(&mut app, Action::InsertChar('x'));
pump_until(&mut app, &mut rx, |a| !a.swarm_dashboard.running).await;

handle_action(&mut app, Action::InsertChar('r'));
assert_eq!(app.swarm_dashboard.view, SwarmView::Form);
assert_eq!(field_text(&app, "Prompt"), "the moon");
}

/// Press `r` during a run; true when the run view stayed on screen.
fn handle_action_is_ignored_while_running(app: &mut App) -> bool {
handle_action(app, Action::InsertChar('r'));
app.swarm_dashboard.view == SwarmView::Run
}

#[tokio::test(flavor = "multi_thread")]
async fn stopping_ends_a_run_that_would_otherwise_never_finish() {
let dir = tempfile::tempdir().unwrap();
Expand Down
69 changes: 66 additions & 3 deletions crates/gaviero-tui/src/panels/swarm_dashboard.rs
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
//! [`crate::app::swarm_run`]; every change arrives through the event loop as a
//! [`SwarmUpdate`].

use std::collections::HashMap;
use std::path::PathBuf;
use std::time::{Duration, Instant};

Expand Down Expand Up @@ -684,9 +685,54 @@ impl SwarmForm {
}
}

/// Take every value `previous` (this script's last run) was launched
/// with: the Folder / Workflow / Profile choices, the agent and tier rows,
/// and every typed input still on the form. Values for inputs the script
/// no longer has are dropped. Focus lands on Run.
pub fn restore_from(&mut self, previous: &SwarmForm) {
for role in [FieldRole::Folder, FieldRole::Workflow, FieldRole::Profile] {
let Some(name) = previous
.fields
.iter()
.find(|f| f.role == role)
.and_then(FormField::choice)
else {
continue;
};
if let Some(field) = self.fields.iter_mut().find(|f| f.role == role)
&& let FieldValue::Choice { options, selected } = &mut field.value
&& let Some(index) = options.iter().position(|o| o == name)
{
*selected = index;
}
}
self.profile_info = previous.profile_info.clone();
let (folder, workflow, profile) = (
self.selected_folder(),
self.selected_workflow(),
self.choice_of(FieldRole::Profile),
);
let mut fresh = self.build_fields(folder, workflow, profile, &previous.fields);
for field in &mut fresh {
if matches!(field.value, FieldValue::Choice { .. }) {
continue;
}
if let Some(old) = previous
.fields
.iter()
.find(|o| o.name == field.name && o.role == field.role && o.is_edited())
{
field.value = old.value.clone();
}
}
self.fields = fresh;
self.focus = self.fields.len();
}

/// Rebuild every field from the current choices, keeping what the user
/// typed or chose under the same name.
fn rebuild(&mut self) {
let on_run = self.on_run_button();
let focused = self.focus;
let folder = self.selected_folder();
let workflow = self.selected_workflow();
Expand All @@ -705,7 +751,11 @@ impl SwarmForm {
}
}
self.fields = fresh;
self.focus = focused.min(self.fields.len());
self.focus = if on_run {
self.fields.len()
} else {
focused.min(self.fields.len())
};
}

/// The first field to type into. The Folder / Workflow / Profile
Expand Down Expand Up @@ -1441,6 +1491,11 @@ pub struct SwarmDashboardState {
pub table_rect: Rect,
pub detail_rect: Rect,
wrapped: WrappedOutput,
/// The form each script was last launched with, by script path, so the
/// next run of that script starts from the same values (this session).
pub last_forms: HashMap<PathBuf, SwarmForm>,
/// Script of the run on screen, for `r`.
pub last_script: Option<PathBuf>,
}

impl SwarmDashboardState {
Expand Down Expand Up @@ -1474,6 +1529,8 @@ impl SwarmDashboardState {
table_rect: Rect::default(),
detail_rect: Rect::default(),
wrapped: WrappedOutput::default(),
last_forms: HashMap::new(),
last_script: None,
}
}

Expand All @@ -1494,7 +1551,13 @@ impl SwarmDashboardState {
self.running = true;
self.abort = None;
self.view = SwarmView::Run;
self.form = None;
// Kept, so running the script again starts from these values.
if let Some(mut form) = self.form.take() {
form.completion.reset();
form.error = None;
self.last_script = Some(form.script.path.clone());
self.last_forms.insert(form.script.path.clone(), form);
}
self.script_label = script_label;
self.run_root = Some(run_root);
self.agents = planned
Expand Down Expand Up @@ -2285,7 +2348,7 @@ impl SwarmDashboardState {
} else if self.running {
"↑↓ agent Enter expand Tab scroll output x stop"
} else {
"Esc scripts ↑↓ agent Enter expand d diff u undo"
"Esc scripts r run again ↑↓ agent Enter expand d diff u undo"
}
}

Expand Down
Loading