Edit Mode and Configure OpenAi Chat Completions

.env.example

@@ -3,3 +3,10 @@ ANTHROPIC_API_KEY=your_anthropic_key_here
 LLM_PROVIDER=deepseek
 DEEPSEEK_API_KEY=
 GROQ_API_KEY=
+
+# Any OpenAI-compatible endpoint (LM Studio, vLLM, llama.cpp server, Together, Azure OpenAI, ...)
+# Set LLM_PROVIDER=openai-compatible and point OPENAI_BASE_URL at the API root.
+# The app appends /chat/completions unless the URL already ends with it.
+OPENAI_BASE_URL=
+OPENAI_COMPATIBLE_MODEL=
+OPENAI_COMPATIBLE_API_KEY=

.github/workflows/release.yml

@@ -1,12 +1,61 @@
 name: Release App
 
 on:
+  pull_request:
+    branches:
+      ["main"]
   push:
+    branches:
+      ["main"]
     tags:
       - 'v*' # Triggers on release tags like v0.1.0 or v1.0.0
 
 jobs:
+  test:
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      # jsdom 30 pulls in undici 8, which needs Node >= 22.19 (it uses
+      # zlib.createZstdDecompress, absent on Node 20).
+      - name: Install Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@stable
+
+      - name: Cache Rust build artifacts
+        uses: Swatinem/rust-cache@v2
+        with:
+          workspaces: src-tauri
+
+      - name: Install Linux dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libwebkit2gtk-4.1-dev build-essential curl wget file libxdo-dev libssl-dev libayatana-appindicator3-dev librsvg2-dev
+
+      - name: Install node modules
+        run: npm ci
+
+      # `tauri::generate_context!` validates `frontendDist: "../dist"`, so the
+      # frontend bundle has to exist before the Rust tests are compiled.
+      - name: Build frontend bundle
+        run: npm run build
+
+      - name: Run frontend tests (Vitest)
+        run: npm test
+
+      - name: Run Rust tests (unit + SQLite integration)
+        run: cargo test --manifest-path src-tauri/Cargo.toml --all-targets
+
   release:
+    # Only real version tags cut a release. On a push to `main` `github.ref_name`
+    # is literally "main", which would create a bogus `main` tag/release.
+    if: startsWith(github.ref, 'refs/tags/v')
+    needs: test
     permissions:
       contents: write
     strategy:
@@ -21,7 +70,8 @@ jobs:
       - name: Install Node.js
         uses: actions/setup-node@v4
         with:
-          node-version: 20
+          node-version: 22
+          cache: npm
 
       - name: Install Rust
         uses: dtolnay/rust-toolchain@stable
@@ -33,6 +83,14 @@ jobs:
           sudo apt-get update
           sudo apt-get install -y libwebkit2gtk-4.1-dev build-essential curl wget file libxdo-dev libssl-dev libayatana-appindicator3-dev librsvg2-dev
 
+      # `bun.lock` used to be committed, which made tauri-action pick bun and fail
+      # with "'bun' is not recognized". package.json now pins npm via packageManager.
+      - name: Assert npm is the package manager
+        shell: bash
+        run: |
+          test ! -f bun.lock && test ! -f bun.lockb
+          node -e "const pm=require('./package.json').packageManager||''; if(!pm.startsWith('npm@')) { console.error('packageManager must pin npm, got: '+pm); process.exit(1); }"
+
       - name: Install node modules
         run: npm ci
 
@@ -41,8 +99,18 @@ jobs:
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         with:
+          # Pinned explicitly: tauri-action picks the runner from lockfiles, and a
+          # committed `bun.lock` made it run `bun tauri build` on runners without bun.
+          tauriScript: npm run tauri
           tagName: ${{ github.ref_name }}
           releaseName: 'AutoShorts ${{ github.ref_name }}'
-          releaseBody: 'See release notes for installation details.'
-          releaseDraft: true
+          # `releaseDraft: true` left every build stuck as a draft that was never
+          # published. Publish directly and let GitHub generate the notes from
+          # the commit history.
+          releaseDraft: false
           prerelease: false
+          generateReleaseNotes: true
+
+
+
+

.gitignore

@@ -5,3 +5,7 @@ src-tauri/target
 .env
 .DS_Store
 *.log
+artifacts
+# Project builds with npm; a stray bun lockfile makes tauri-action pick bun.
+bun.lock
+bun.lockb

.vscode/settings.json

@@ -0,0 +1,4 @@
+{
+    "python-envs.defaultEnvManager": "ms-python.python:poetry",
+    "python-envs.defaultPackageManager": "ms-python.python:poetry"
+}

README.md

@@ -88,7 +88,7 @@ When you first launch the application, you will be greeted by an **Onboarding Wi
 
 #### Option A: Fully Offline (Ollama + Whisper)
 1. **Ollama Setup**: Select a local model card (`llama3.2 3B`, `qwen2.5 3B`, or `qwen2.5 7B`). The application will check if Ollama is running and automatically pull the model weights, showing a downloader progress bar.
-2. **Local Whisper**: Follow the prompt instructions to verify Python is installed and run `pip3 install openai-whisper` to enable fully offline transcription.
+2. **Local Whisper**: Click **Instalar Whisper**. The app creates a virtualenv under its data directory (`python-venv`) and installs `openai-whisper` into it, so nothing is added to your system Python. The first install downloads PyTorch (~1–2 GB). Only a Python 3 interpreter is required — the Linux packages declare `python3`/`python3-venv` as dependencies.
 
 #### Option B: Cloud API Keys
 1. Enter your API credentials for:
@@ -97,6 +97,14 @@ When you first launch the application, you will be greeted by an **Onboarding Wi
    * **Claude**: For premium copywriting, hooks, and moment detection.
 2. Click **Save & Start** to immediately load the dashboard.
 
+#### Option C: Any OpenAI-Compatible Endpoint
+Select **OpenAI-Compatible (Custom Base URL)** in **API Settings** and provide:
+* **Base URL**: the API root, e.g. `http://localhost:1234/v1` (LM Studio), `http://localhost:8000/v1` (vLLM), `http://localhost:8080/v1` (llama.cpp), or a hosted provider URL.
+* **Model Name**: the model identifier that endpoint expects.
+* **API Key**: optional; leave empty for local servers that do not require auth.
+
+The app calls `POST {Base URL}/chat/completions` and retries without `response_format` if the server rejects JSON mode.
+
 ---
 
 ### ⚙️ Modifying Settings & Resetting Onboarding
@@ -125,8 +133,14 @@ DEEPGRAM_API_KEY=your-deepgram-api-key
 DEEPSEEK_API_KEY=your-deepseek-api-key
 ANTHROPIC_API_KEY=your-anthropic-api-key
 
-# Choose your default AI analysis provider ("deepseek" or "claude")
+# Choose your default AI analysis provider
+# ("deepseek", "claude", "gemini", "openai", "openrouter", "groq", "local" or "openai-compatible")
 LLM_PROVIDER=deepseek
+
+# Only for LLM_PROVIDER=openai-compatible
+OPENAI_BASE_URL=http://localhost:1234/v1
+OPENAI_COMPATIBLE_MODEL=qwen2.5-7b-instruct
+OPENAI_COMPATIBLE_API_KEY=
 ```
 
 ### 2. Run in Development Mode
@@ -143,6 +157,62 @@ npm run tauri:build
 ```
 The output installers will be built under `src-tauri/target/release/bundle/`.
 
+### 4. Run the Tests
+Rust unit tests live next to the code they cover (`#[cfg(test)]` modules in
+`src-tauri/src/*.rs`); the SQLite layer is covered end to end from
+`src-tauri/tests/db_integration.rs` against a temporary database file:
+
+```bash
+cargo test --manifest-path src-tauri/Cargo.toml
+# or
+npm run test:rust
+```
+
+Frontend tests use Vitest + Testing Library with a jsdom environment:
+
+```bash
+npm test              # single run
+npm run test:watch    # watch mode
+```
+
+Test layout:
+
+| Path | Kind | Covers |
+| --- | --- | --- |
+| `src-tauri/src/lib.rs` | unit | SRT/ASS caption generation, slug + extension validation |
+| `src-tauri/src/media.rs` | unit | ffmpeg progress parsing, ASS filter escaping, path handling |
+| `src-tauri/src/transcription.rs` | unit | Whisper/Deepgram normalization, segment building, download progress |
+| `src-tauri/src/llm.rs` | unit | LLM response parsing, SSE streaming, URL normalization |
+| `src-tauri/tests/db_integration.rs` | integration | projects, transcripts, candidates, clips, cascade deletes, migrations |
+| `src/lib/format.test.ts` | unit | duration/size/file-name formatting helpers |
+| `src/components/common.test.tsx` | component | `PipelineStep`, `EmptyState`, `StatusPill` |
+| `src/App.integration.test.tsx` | integration | onboarding gate, sidebar rendering, project detail loading (Tauri IPC mocked) |
+
+### 5. Multi-Platform Build (Windows + Linux + macOS)
+One command attempts every platform the current host can actually target:
+```bash
+npm run build:all
+```
+Each platform is detected first; unsupported ones are skipped with the reason instead of aborting the run. Successful bundles are copied to `artifacts/<platform>/`.
+
+```bash
+npm run build:linux     # .deb / .rpm / .AppImage
+npm run build:windows   # .msi / NSIS .exe
+npm run build:macos     # .dmg / .app (macOS only)
+
+npm run build:all -- --only=linux,macos     # subset
+npm run build:all -- --skip=windows         # exclude one
+npm run build:all -- --dry-run              # print commands only
+```
+
+Host capability matrix:
+
+| Target | Requires |
+| --- | --- |
+| Linux | Linux host + `webkit2gtk-4.1`, `gtk+-3.0`, `libsoup-3.0` (pkg-config) |
+| Windows | Windows host, **or** WSL with Windows `cargo.exe` + `node.exe` + MSVC Build Tools. From WSL the source is mirrored to `%USERPROFILE%\autoshorts-build` and built by the Windows toolchain. |
+| macOS | macOS + Xcode only. Cannot be produced from Linux/WSL — use a Mac or the `Release App` GitHub workflow. |
+
 <a id="support"></a>
 
 ## ❤️ Support AutoShorts