Skip to content

development.md

Latest commit

 

History

History
129 lines (101 loc) · 5.04 KB

development.md

File metadata and controls

129 lines (101 loc) · 5.04 KB
Raw

Development Guide

Stack: Svelte 5 + Tauri 2 + Rust · Build: Vite (frontend) + Cargo (backend)

1. Prerequisites

Tool Version Check Get it
Node.js ≥ 18 node --version nodejs.org
npm ≥ 9 npm --version (ships with Node)
Rust / Cargo ≥ 1.77 rustc --version rustup.rs
WebView2 — pre-installed on Win 10 21H2+ / Win 11 Microsoft
ffmpeg / ffprobe — ffmpeg -version on PATH or next to the exe (video only)
LM Studio any — lmstudio.ai (or any OpenAI-compatible vision server)

The Tauri CLI is provided through package.json (@tauri-apps/cli); no global install is required — use npx tauri ….

2. First-time setup

git clone <repo-url>
cd meta-analyzer
npm install

# sanity check
npm run build                                        # frontend (Vite)
cargo check --manifest-path src-tauri/Cargo.toml     # backend (Rust)

# run with hot-reload
npx tauri dev

3. Everyday workflow

Task Command Notes
Dev with hot-reload npx tauri dev Frontend hot-reloads instantly; Rust changes trigger a rebuild.
Debug build + console .\build-debug.ps1 Runs the debug exe attached to PowerShell so you see println! / eprintln! and backtraces.
Release (installer + portable) .\build-release.ps1 Produces the NSIS installer and the portable package into .\release.
Frontend only npm run build Output to dist/.
Backend only (release) cargo build --release --manifest-path src-tauri/Cargo.toml Optimized meta-analyzer.exe.

Because the repo root is a Cargo workspace, target/ is created at the repo root (not inside src-tauri/).

4. Where things live

What Path
Frontend entry src/main.ts
Root component src/App.svelte
UI components src/components/*.svelte
Global state src/lib/store.svelte.ts
Shared types src/lib/types.ts
IPC bridge src/lib/tauri.ts
Rust entry src-tauri/src/main.rs
Commands src-tauri/src/commands/*.rs
Processing src-tauri/src/processing/*.rs
Content rules src-tauri/content_rules/*.json

5. Adding things

A Tauri command

  1. Write the function in src-tauri/src/commands/…, annotate with #[tauri::command].
  2. Register it in the invoke_handler![...] list in main.rs.
  3. Add a wrapper in src/lib/tauri.ts.
  4. Call it from the store or a component.

An event

  1. Define a #[derive(Serialize)] payload struct (in models.rs / video_models.rs).
  2. Emit it with app.emit("event-name", payload)? (use tauri::Emitter).
  3. Add the TS interface in types.ts and a listener in tauri.ts.
  4. Subscribe in App.svelte's onMount.

A content category Drop an image_*.json / video_*.json rule file into the content-rules folder — it is discovered at runtime, no code change needed. See prompt-builder.md.

6. Rust conventions

  • Commands return Result<T, String>. Convert errors with .map_err(|e| e.to_string()) / format!.
  • Serde uses camelCase (#[serde(rename_all = "camelCase")]) so the frontend always sees/sends camelCase.
  • Emit via &AppHandle / &Window, never a global.
  • Shared state (ProcessingQueue, VideoProcessingQueue) is registered with .manage() and pulled into commands via State<'_, T>.
  • Data paths always go through paths::data_dir(app) — never app.path().app_data_dir() directly — so portable mode keeps working.

7. Debugging

# build-debug.ps1 sets these; or manually:
$env:RUST_LOG = "info,meta_analyzer=debug,tauri=info"
$env:RUST_BACKTRACE = "1"
.\target\debug\meta-analyzer.exe

Frontend: console.log shows in the WebView console (right-click → Inspect in dev builds). Release builds have DevTools disabled.

Symptom Likely cause / fix
Failed to fetch models LLM server not running / wrong API URL.
No tags parsed Model returned an unexpected shape — try a stronger model.
Metadata write failed File read-only or in use.
No package info in the config file (build) A stray tauri.conf.json outside src-tauri/, or a stale target/.

8. Model setup (local)

  1. Install LM Studio, load a vision model, start its local server (default http://localhost:1234/v1).
  2. In the app, set the API URL, click refresh, pick the model.
  3. Reasoning-capable models produce noticeably more reliable JSON. See the README for concrete model recommendations.

9. Release process

# bump version in: src-tauri/Cargo.toml, package.json, src-tauri/tauri.conf.json
.\build-release.ps1
# artifacts land in .\release (installer + portable + zip)

There are currently no automated tests; verify changes manually (import → process → apply → check written metadata with e.g. ExifTool / a media player).