6 round 3 collections workspaces sidebar tree tabs file persistence v2

.claude/agents/senior-rust-engineer.md

@@ -7,6 +7,26 @@ model: sonnet
 
 You are a senior Rust engineer with deep expertise in TUI development using ratatui and async programming with tokio. You are building **forge** — a terminal-native API client (Postman in the terminal). Read `SPEC.md` for the full spec and `CLAUDE.md` for project conventions before writing code.
 
+---
+
+## Pre-Implementation Reasoning (MANDATORY)
+
+**Before writing any code**, run the reasoning skill checklist at `.claude/skills/reasoning/SKILL.md`.
+
+Work through every section that applies to your task:
+
+1. **Data Model** — new/changed structs, enum variants, serde defaults, exhaustive matches
+2. **Lifecycle/Sync** — when created, modified, flushed, destroyed, loaded on restart
+3. **Idempotency/Dedup** — what if triggered twice? search before push
+4. **Inverse Operations** — every open/create needs a matching close/delete/save
+5. **UI Completeness** — focus cycle, keybind hints, empty state, overflow/scroll
+6. **State Coherence** — clamp indices after mutations, invalidate caches
+7. **"Who Else Touches This?" Audit** — grep every symbol you change; categorize by create/read/update/delete/persist/display
+
+Answer each relevant question (one-liner is enough) before touching a single source file. Unanswered questions = gaps that will become bugs.
+
+---
+
 ## Core Principles
 
 - **Correctness first**: safe Rust, no `unwrap()` in production paths — use `?` and proper error types

.claude/settings.local.json

@@ -6,7 +6,10 @@
       "Bash(echo No tokio-util in lock yet:*)",
       "WebFetch(domain:docs.rs)",
       "Bash(cd \"C:\\\\Users\\\\alber\\\\Desktop\\\\forge\" && cargo check 2>&1)",
-      "Bash(ls:*)"
+      "Bash(ls:*)",
+      "Bash(xargs grep:*)",
+      "Bash(cargo check:*)",
+      "Bash(cargo test:*)"
     ]
   },
   "enableAllProjectMcpServers": true,

.claude/skills/reasoning/SKILL.md

@@ -0,0 +1,134 @@
+---
+name: reasoning
+description: >
+  Self-questioning checklist to run before implementing any feature.
+  Use to surface edge cases, lifecycle gaps, persistence requirements,
+  dedup needs, and UI completeness issues before writing code.
+  Invoke mentally (or literally write answers) before starting implementation.
+---
+
+# Feature Reasoning Checklist
+
+Before writing code, work through these question groups out loud.
+Answer each one briefly — a one-liner is enough. Unanswered questions = gaps in the plan.
+
+---
+
+## 1. Data Model Questions
+
+*Trigger: adding or changing any struct field, enum variant, or type.*
+
+- Does this new field need to be **persisted** (saved to disk / TOML)?
+  → If yes: add `#[serde(default)]` so old files still load.
+- Are there **struct literal initializations** (`Struct { field: val, ... }`) elsewhere
+  that will now fail to compile?
+  → Search for every place the struct is constructed by name.
+- Does the `Default` impl need updating?
+- Do any `Clone`, `PartialEq`, or `Display` impls need to handle the new field?
+- If I added an enum variant: are there `match` statements on this enum elsewhere?
+  → Every exhaustive match must get a new arm.
+
+---
+
+## 2. Lifecycle / Sync Questions
+
+*Trigger: any feature that creates, modifies, or destroys stateful data.*
+
+- **When is this data created?** (startup, user action, event?)
+- **When is it modified?** (every keystroke, on submit, on navigate-away?)
+- **When should it be flushed/saved?** Never assume "it's saved automatically."
+  → List every code path that changes it; each path needs a save call or a sync point.
+- **When is it destroyed?** (tab close, workspace switch, app exit?)
+  → Every destruction path must flush state first.
+- What happens on **app restart**? Is the state loaded back correctly?
+  → Test: create → close app → reopen → is state intact?
+
+---
+
+## 3. Idempotency / Dedup Questions
+
+*Trigger: any feature that opens, creates, or adds something.*
+
+- What happens if the user **triggers this action twice** in a row?
+  → Should it be a no-op, an error, or create a second item?
+- Is there an **existing open instance** that should be re-focused instead?
+  → Search open_tabs / active list before pushing a new item.
+- What if the **item already exists** with the same identity?
+  → Define identity: is it by ID, by name, by position?
+
+---
+
+## 4. Inverse / Symmetric Operations
+
+*Trigger: any "open", "create", "start", or "enable" action.*
+
+For every action X, list its inverses and verify each handles the new state:
+
+| Action added | Inverses to check |
+|---|---|
+| open tab | close tab, switch tab, switch workspace, app exit |
+| create item | delete item, rename item, duplicate item |
+| add field | all constructors, all serialization round-trips |
+| enable feature | disable feature, reset to default |
+
+**Every entry point needs a matching exit point that cleans up / persists.**
+
+---
+
+## 5. UI Completeness Questions
+
+*Trigger: any new visible element (widget, panel, row, popup, indicator).*
+
+- Should this element be **keyboard-navigable**?
+  → If yes: add a `Focus` variant, update `next()`/`prev()`, add visual indicator.
+- Does it need a **direct shortcut key** (number, letter)?
+- Does it need **keybinding hints** in a status bar or hint footer?
+- What happens when the element is **empty** (zero items, blank state)?
+  → Render a placeholder; don't panic on `list[0]`.
+- What happens when the element is **out of screen bounds** (many items, small terminal)?
+  → Clamp scroll_offset; ensure cursor stays visible.
+
+---
+
+## 6. Related State Coherence
+
+*Trigger: any action that adds or removes items from a list, or changes active indices.*
+
+- After this action, are all **indices still valid**?
+  → active_tab_idx, cursor, scroll_offset — clamp them after mutations.
+- Are there **other state fields** that reference the mutated data by index or ID?
+  → Update or invalidate caches (e.g. highlighted_body, selected row).
+- Does any **other component render** based on the data I changed?
+  → Read render functions that touch the same state; ensure they handle new shape.
+
+---
+
+## 7. The "Who Else Touches This?" Audit
+
+*Run this for every struct, field, or function you modify.*
+
+1. Search the codebase for all uses of the symbol.
+2. Categorize: create / read / update / delete / display / persist / test.
+3. For each category: does my change break or require updating that site?
+
+```
+Symbol: CollectionRequest
+  create: CollectionRequest::new(), struct literal in sidebar_duplicate → needs url/body_raw
+  read:   flatten_tree(), find_col_request_by_id() → fine, reads by field
+  update: update_col_request_state() → needs url/body_raw
+  persist: save_collection_meta() → handled by serde
+  open:   handle_sidebar_enter() → needs to load url/body_raw back
+  close:  close_active_tab() → needs to sync url/body_raw first
+```
+
+---
+
+## Checklist (run before every implementation)
+
+- [ ] Listed all struct literals for modified structs → all compile?
+- [ ] Named every lifecycle stage (create / modify / destroy) and wired save/load
+- [ ] Checked for dedup: what if this is triggered twice?
+- [ ] Verified all inverses (close, delete, switch) handle the new state
+- [ ] If new UI element: added to Focus cycle, visual indicator, hint bar
+- [ ] Indices/cursors clamped after any list mutation
+- [ ] "Who else touches this?" audit complete — no missed call sites

PROGRESS.md

@@ -4,7 +4,7 @@
 
 - [x] Round 1 — Core Request Engine
 - [x] Round 2 — Environment Variables (partly - [more details here](SPEC.md#implementation-tasks-1))
-- [ ] Round 3 — Collections & Workspaces
+- [x] Round 3 — Collections & Workspaces
 - [ ] Round 4 — Authentication
 - [ ] Round 5 — Request Headers & Query Params (Done Partly)
 - [ ] Round 6 — Request Body Editor (Done Partly)
@@ -124,6 +124,51 @@
 
 ---
 
+## Round 3 — Collections & Workspaces ✓
+
+### Files Implemented (13 new/major files)
+
+| Layer | Files |
+|---|---|
+| State | `state/collection.rs`, `state/workspace.rs`, `state/app_state.rs` (major migration) |
+| Storage | `storage/workspace.rs`, `storage/collection.rs` |
+| UI | `ui/sidebar.rs` (full rewrite), `ui/request_tabs.rs`, `ui/naming_popup.rs`, `ui/confirm_delete.rs`, `ui/workspace_switcher.rs` |
+| App Logic | `app.rs` (sidebar CRUD, tab management, workspace switching) |
+
+### Architecture
+
+- **AppState migration**: `state.request`/`state.response` → `state.workspace.open_tabs[active_tab_idx]`; accessed via `state.active_tab()` / `state.active_tab_mut()`
+- **Environments moved**: `state.environments` → `state.workspace.environments`; `state.active_env_idx` → `state.workspace.active_environment_idx`
+- **Storage path**: `%APPDATA%/forge/workspaces/<ws-name>/` (Windows) / XDG / macOS equivalents
+- **Sidebar tree**: `SidebarNode` enum flattened to a list via `flatten_collections()`; collapsed node IDs tracked in `sidebar.collapsed_ids: HashSet<Uuid>`
+- **Tabs**: `WorkspaceState.open_tabs: Vec<RequestTab>`; `active_tab_idx` tracks focus; tabs persist to `workspace.toml`
+
+### Keybindings Added
+
+| Key | Action |
+|---|---|
+| `Ctrl+W` | Workspace switcher popup |
+| `Ctrl+n` (Sidebar) | New collection |
+| `n` (Sidebar) | New request |
+| `f` (Sidebar) | New folder |
+| `r` (Sidebar) | Rename selected item |
+| `d` (Sidebar) | Delete selected item |
+| `D` (Sidebar) | Duplicate selected item |
+| `h` / `l` (Sidebar) | Collapse / expand node |
+| `/` (Sidebar) | Toggle search mode |
+| `Alt+1–9` | Switch to tab N |
+| `Alt+w` | Close active tab |
+| `[` / `]` | Cycle open tabs (non-UrlBar focus) |
+
+### Gotchas & Fixes
+
+- Sidebar search runs inline at the footer row (repurposed hint row); `NamingState` carries the HTTP method for new requests so method persists through the naming popup flow
+- `active_tab()` returns `Option<&RequestTab>` — all render functions must handle `None` gracefully
+- Workspace save triggered on every tab/request mutation via `dirty` flag; debounced via the Tick event
+- `flatten_collections()` recurses into folders and respects `collapsed_ids` to hide children
+
+---
+
 ## Round 5 — Request Headers & Query Params (Done Partly)
 
 ### What's Implemented

README.md

@@ -95,7 +95,7 @@ Every request, collection, environment, and workspace is a human-readable TOML f
 
 - [x] **Round 1** - Core Request Engine (URL bar, HTTP executor, response viewer, syntax highlighting)
 - [x] **Round 2** - Environment Variables (`{{variable}}` interpolation, env switcher, secret vars) (partly - [more details here](SPEC.md#implementation-tasks-1))
-- [ ] **Round 3** - Collections & Workspaces (sidebar tree, tabs, file persistence)
+- [x] **Round 3** - Collections & Workspaces (sidebar tree, tabs, file persistence)
 - [ ] **Round 4** - Authentication (Basic, Bearer, API Key, OAuth 2.0, Digest)
 - [ ] **Round 5** - Request Headers & Query Params (key-value editors, autocomplete, bidirectional URL sync)
 - [ ] **Round 6** - Request Body Editor (JSON, Form, Multipart, GraphQL, Raw, Binary)

SPEC.md

@@ -908,15 +908,15 @@ pub struct RequestTab {
 
 ### Implementation Tasks
 
-- [ ] Implement sidebar tree widget with collapse/expand
-- [ ] Implement collection CRUD (new, rename, delete, duplicate)
-- [ ] Implement folder support inside collections
-- [ ] Implement request tabs in main panel
-- [ ] Implement workspace switcher popup
-- [ ] Implement sidebar fuzzy search
-- [ ] Implement file persistence for collections (TOML files)
-- [ ] Implement auto-save on request modification
-- [ ] Implement workspace-level CRUD
+- [x] Implement sidebar tree widget with collapse/expand
+- [x] Implement collection CRUD (new, rename, delete, duplicate)
+- [x] Implement folder support inside collections
+- [x] Implement request tabs in main panel
+- [x] Implement workspace switcher popup
+- [x] Implement sidebar fuzzy search
+- [x] Implement file persistence for collections (TOML files)
+- [x] Implement auto-save on request modification
+- [x] Implement workspace-level CRUD
 - [ ] Display unsaved indicator (`*`) on dirty requests/tabs
 
 ---