Events, async watch rework, bug fixes, enhancements (#199)

Signed-off-by: Daniel Guns <danbguns@gmail.com>

Author: dgunzy

CHANGELOG.md

@@ -8,6 +8,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [Unreleased]
 
 ### Added
+- Kubernetes Events support:
+  - The describe view (`d`) now ends with a kubectl-style Events section
+    showing the resource's recent events (Warnings highlighted; degrades to a
+    notice when events are unavailable, e.g. RBAC).
+  - New `:events` (alias `:ev`) live events feed for the current namespace
+    scope — cluster-wide with `:ns all`. Streams in real time, newest first,
+    filterable with `/`; `Enter` jumps to the involved Flux resource. The
+    events watcher runs only while the view is open.
+- Resource action keys now resolve their target from any view via a single
+  `view_target()` resolver: `y`/`d` (and `t`/`g`/`h`/operations for watched
+  resources) act on the selected event's involved object in the events feed
+  and on the focused node in the graph view — previously these keys only
+  worked from the resource list, favorites, and detail views. `y`/`d` work
+  for non-Flux objects too (Pods, Deployments, …) via the API fetch path,
+  and Back returns to the view you came from (events feed or graph).
 - Graph view keyboard navigation: `j`/`k` move a highlighted focus between nodes
   (the view auto-scrolls to keep it visible), `Enter` opens the focused resource's
   detail view, and `Esc` returns to the graph.
@@ -17,7 +32,23 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   children → drop into each child) using proper box-drawing junctions, and nodes
   sit closer together for a tighter, clearer layout.
 
+### Fixed
+- `:all` now returns to the main resource list from the events feed (and stops
+  the events watch), instead of only clearing filters while stuck in the view.
+- The events feed's "not a watched Flux resource" message now names the
+  involved object's namespace and points at `y`/`d`, so namespace-scope
+  mismatches are visible instead of just confusing.
+
 ### Internal
+- New generic `AsyncTask<K, T>` owns the request/dispatch/poll lifecycle for
+  view fetches, replacing the five copy-pasted `*_pending`/`*_fetched`/`*_rx`
+  field triplets and their hand-rolled trigger/poll methods.
+- YAML/describe fetch requests are now typed `ResourceKey`s instead of
+  `"type:namespace:name"` strings, removing the re-parse (and its can't-happen
+  error branches) from the main loop.
+- The Kubernetes events watcher has an independent lifecycle from the resource
+  watchers (started lazily, stopped without a full watcher teardown), and
+  survives namespace switches.
 - Single source of truth for graph node sizing (`GraphNode::render_width`/
   `render_height`); connector geometry extracted into the testable, `Frame`-free
   `fanout_routes()`.

Cargo.lock

@@ -106,10 +106,15 @@ Terminal user interface built with ratatui.
 
 - **`app/`** - Application state and logic (refactored from single file):
   - `core.rs` - Main App struct and core logic
-  - `state.rs` - State structures (ViewState, SelectionState, UIState, AsyncOperationState)
+  - `state.rs` - State structures (ViewState, SelectionState, UIState, AsyncOperationState, KubeEventStore)
   - `events.rs` - Event handling and input processing
   - `rendering.rs` - Rendering orchestration
-  - `async_ops.rs` - Async operation management
+  - `async_task.rs` - Generic `AsyncTask<K, T>`: one type owns the
+    request → dispatch → poll lifecycle for every view fetch (YAML, describe,
+    trace, graph). Event handlers call `request()`, the main loop calls
+    `dispatch()` and spawns the work, then polls `try_recv()` each tick.
+  - `async_ops.rs` - Mutating-operation flow (registry validation, status
+    message bookkeeping) and the graph result hook
 - **`operations.rs`** - Flux operations (suspend, resume, delete, reconcile, reconcile with source)
 - **`theme.rs`** - Theme configuration and loading
 - **`trace.rs`** - Trace operation orchestration
@@ -127,6 +132,7 @@ Terminal user interface built with ratatui.
   - `trace.rs` - Trace view showing resource ownership chains
   - `graph.rs` - Graph visualization view
   - `history.rs` - Reconciliation history view
+  - `events.rs` - Live Kubernetes events feed (`:events`)
   - `confirmation.rs` - Confirmation dialogs
   - `help.rs` - Help screen
   - `splash.rs` - Splash screen
@@ -139,6 +145,10 @@ Terminal user interface built with ratatui.
 - Non-blocking async operations using `tokio::spawn`
 - Modular app structure with separated concerns (state, events, rendering, async)
 - Separate scroll offsets for different views
+- The Kubernetes events watcher (`ResourceWatcher::watch_kube_events`) has an
+  independent lifecycle from the resource watchers: started lazily when the
+  events view opens, stopped (and its `KubeEventStore` feed cleared) when the
+  view is left, and restarted automatically across namespace/context switches
 - Per-view behavior (scroll target, back navigation, classification) lives on
   `impl View` (`src/tui/app/state.rs`) rather than scattered `match` arms
 - Dynamic footer wrapping for smaller screens

DEVELOPER_GUIDE.md

@@ -31,6 +31,7 @@ A [K9s](https://github.com/derailed/k9s)-inspired terminal UI for monitoring Flu
 - **Unified and type-specific views** - View all resources together or filter by type
 - **Resource operations** - Suspend, resume, reconcile, and delete Flux resources
 - **YAML viewing** - Inspect full resource manifests
+- **Kubernetes Events** - Per-resource events in the describe view, plus a live `:events` feed for the current namespace or cluster
 - **Graph visualization** - Visualize resource relationships and dependencies (Kustomization, HelmRelease, etc.)
 - **Reconciliation history** - View reconciliation history for resources that track it
 - **Favorites** - Mark frequently accessed resources for quick access
@@ -143,6 +144,7 @@ By default, `flux9s` watches the `flux-system` namespace. Use `:ns all` to view
 - `:ns <namespace>` - Switch namespace
 - `:ns all` - View all namespaces
 - `:favorites` or `:fav` - View favorite resources
+- `:events` or `:ev` - Live Kubernetes events feed (current namespace scope)
 - `:skin {skin-name}` - set skin directly
 - `:skin` - open interactive theme selection menu with live preview (17 built-in themes + custom)
 - `:q` or `:q!` - Quit
@@ -152,6 +154,7 @@ By default, `flux9s` watches the `flux-system` namespace. Use `:ns all` to view
 
 - **Graph View (`g`)** - Visualize resource relationships and dependencies. Shows upstream sources and downstream managed resources. Move the highlighted focus between nodes with `j`/`k` (the view scrolls to keep it visible), press `Enter` to open the focused resource's detail view, and `Esc` to return to the graph. Supported for Kustomization, HelmRelease, ArtifactGenerator, FluxInstance, and ResourceSet.
 - **History View (`h`)** - View reconciliation history for FluxInstance, ResourceSet, Kustomization, and HelmRelease resources.
+- **Events View (`:events`)** - Live Kubernetes events feed for the current namespace (or cluster-wide with `:ns all`), newest first with Warnings highlighted. `Enter` jumps to the involved resource; the describe view (`d`) also shows a per-resource Events section.
 - **Favorites (`f`)** - Mark resources as favorites for quick access. Use `:favorites` command to view all favorites.
 
 ### Terminal Commands

README.md

@@ -95,6 +95,8 @@ Type these commands in command mode (press `:`):
 | `:unhealthy`       | Show only unhealthy resources            |
 | `:favorites`       | View favorite resources                  |
 | `:fav`             | Alias for `:favorites`                   |
+| `:events`          | Live Kubernetes events feed              |
+| `:ev`              | Alias for `:events`                      |
 | `:skin <name>`     | Change theme/skin (direct)               |
 | `:skin`            | Open interactive theme selection menu    |
 | `:readonly`        | Toggle readonly mode                     |
@@ -219,7 +221,8 @@ The graph view displays:
 
 - `j` / `k` (or `↓` / `↑`) - Move the highlighted focus between nodes; the view scrolls to keep the focused node visible.
 - `Enter` - Open the focused node's resource in the detail view. Aggregate nodes (workload/resource groups) and external upstream URLs aren't directly openable.
-- `Esc` / `Backspace` - Return to the graph (when you opened a detail view from it), then back to the resource list.
+- `y` / `d` - View the focused node's YAML or describe output directly, including managed workloads (Deployments, Services, etc.).
+- `Esc` / `Backspace` - Return to the graph (when you opened a view from it), then back to the resource list.
 
 Focus starts on the resource you opened the graph from, so you can immediately walk its sources and dependencies.
 
@@ -250,6 +253,31 @@ Mark frequently accessed resources as favorites for quick access.
 - Favorites are saved to your configuration file
 - Favorites appear first in resource lists
 
+### Events View (`:events`)
+
+A live feed of Kubernetes Events in the current namespace scope — the
+"what is Flux doing right now" view. Flux controllers emit an Event for every
+reconciliation success and failure, so this surfaces error detail that the
+resource list's MESSAGE column truncates.
+
+- The feed follows your namespace scope: the current namespace by default, or
+  the whole cluster after `:ns all` (a NAMESPACE column appears)
+- Events are streamed in real time, newest first, with Warnings highlighted
+- `/` filters by type, reason, object, namespace, source, or message text
+- `Enter` on an event jumps to the involved resource's detail view when it is
+  a Flux resource flux9s watches; `Esc` returns to the events feed
+- Resource keys act on the selected event's involved object directly: `y`
+  (YAML) and `d` (describe) work even for non-Flux objects like Pods and
+  Deployments, while `t`/`g`/`h` and operations (`s`/`r`/`R`) work when the
+  object is a watched Flux resource. `Esc` from any of these returns to the
+  events feed
+- `Esc` from the feed returns to the resource list and stops the events watch —
+  events are only streamed while the view is open, so there is no overhead the
+  rest of the time
+
+Events also appear in the describe view (`d`): each resource's describe output
+ends with a kubectl-style Events section listing that resource's recent events.
+
 ## Operations
 
 Perform actions on selected resources:

docs/content/user-guide/_index.md

@@ -9,6 +9,11 @@ pub const RESOURCE_KEY_FORMAT: &str = "resource_type:namespace:name";
 /// Maximum number of reconciliation history events to store per resource
 pub const MAX_RECONCILIATION_HISTORY: usize = 50;
 
+/// Cap on the live Kubernetes events feed. Events are the churniest resource
+/// in a cluster; the store evicts oldest-seen entries past this bound so a
+/// busy cluster can't grow memory without limit.
+pub const MAX_KUBE_EVENTS: usize = 1000;
+
 /// Status message timeout in seconds
 pub const STATUS_MESSAGE_TIMEOUT_SECS: u64 = 4;