fix: Add WASM embedding, shared database location, and CLI/desktop parity

## WASM Embedding (v0.4.1)
- Embed sql-wasm.wasm as base64 for standalone distribution
- Extract to temp file at runtime (/tmp/forest-sql-wasm-<pid>.wasm)
- Fixes ENOENT errors in distributed binaries
- Fallback to development path when running from source

## Shared Database Location (v0.4.2)
- CLI and desktop now use same database: ~/Library/Application Support/com.ettio.forest.desktop/
- Platform-specific paths (macOS/Linux/Windows)
- Override with FOREST_DB_PATH environment variable
- Ensures seamless integration between CLI and GUI

## Embedding Parity (v0.4.3)
- CLI now uses same fastembed engine as desktop app
- forest-embed Rust helper bundled alongside CLI
- Both produce identical 384-dim embeddings (all-MiniLM-L6-v2)
- True feature parity between CLI and desktop

## Other Changes
- Updated .gitignore to exclude large binaries and build artifacts
- Updated documentation with all fixes
- Version bump to 0.4.3

Author: bwl

.claude/settings.local.json

@@ -12,9 +12,15 @@
       "Bash(cargo build:*)",
       "WebFetch(domain:v2.tauri.app)",
       "WebFetch(domain:edwardtufte.github.io)",
-      "WebFetch(domain:docs.rs)"
+      "WebFetch(domain:docs.rs)",
+      "Bash(bun install)",
+      "Bash(bun run build:*)",
+      "WebFetch(domain:docs.brew.sh)"
     ],
     "deny": [],
     "ask": []
-  }
+  },
+  "enabledMcpjsonServers": [
+    "tauri-mcp"
+  ]
 }

.gitignore

@@ -39,6 +39,10 @@ Thumbs.db
 forest-desktop/src-tauri/target/
 forest-desktop/src-tauri/binaries/*.dmg
 forest-desktop/src-tauri/binaries/*.app
+forest-desktop/src-tauri/binaries/*-aarch64-apple-darwin
+forest-desktop/src-tauri/binaries/*-x86_64-apple-darwin
+forest-desktop/src-tauri/binaries/*-x86_64-pc-windows-msvc.exe
+forest-desktop/src-tauri/binaries/*-x86_64-unknown-linux-gnu
 
 # Bun build cache
 .*.bun-build
@@ -49,3 +53,10 @@ forest-desktop/dist/
 # MCP socket files
 *.sock
 .mcp.json
+
+# Embedding model cache
+.fastembed_cache/
+
+# Local Claude settings
+.claude/settings.local.json
+forest-desktop/.claude/settings.local.json

CLAUDE.md

@@ -6,6 +6,12 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 Forest is a graph-native knowledge base CLI that captures unstructured ideas and automatically links them using a hybrid scoring algorithm combining semantic embeddings and lexical similarity. All data is stored in a single SQLite database (`forest.db`).
 
+**Database Location (as of v0.4.2):**
+- **macOS**: `~/Library/Application Support/com.ettio.forest.desktop/forest.db`
+- **Linux**: `~/.local/share/com.ettio.forest.desktop/forest.db`
+- **Windows**: `%APPDATA%\com.ettio.forest.desktop\forest.db`
+- **Override**: Set `FOREST_DB_PATH` environment variable to use a custom location
+
 ## Agent-First TLDR Standard
 
 Forest implements the **TLDR Standard (v0.1)** for agent ingestion - a minimal, parseable command metadata format designed for AI agents.
@@ -382,7 +388,7 @@ Uses **sql.js** (SQLite compiled to WASM) with in-memory database persisted to d
 
 **Key pattern**: Database is lazily initialized on first access. All mutations set `dirty = true` and persist to disk.
 
-Database path controlled by `FOREST_DB_PATH` env var (default: `forest.db` in cwd).
+Database path controlled by `FOREST_DB_PATH` env var (default: app data directory, see above).
 
 **Database Schema:**
 ```sql
@@ -509,14 +515,24 @@ Then applies a 0.9× penalty if both `tagOverlap` and `titleSimilarity` are zero
 
 **Token downweighting**: Generic technical terms (flow, stream, pipe, branch, terminal) are weighted at 0.4× in token similarity to reduce over-connection of unrelated domains.
 
-### Embeddings (src/lib/embeddings.ts)
+### Embeddings
+
+**Desktop App (Rust):** Uses `fastembed` crate with `all-MiniLM-L6-v2` (384-dim) - native, fast, bundled
 
-Three providers via `FOREST_EMBED_PROVIDER`:
-1. **local** (default): Uses `@xenova/transformers` with model `Xenova/all-MiniLM-L6-v2` (384-dim)
+**CLI (TypeScript):** Three providers via `FOREST_EMBED_PROVIDER`:
+1. **local** (default): Uses `forest-embed` Rust helper (same fastembed engine as desktop app!)
+   - Ensures CLI and desktop app produce identical embeddings
+   - Binary located next to CLI executable
+   - Falls back to mock if binary not found
 2. **openai**: Calls OpenAI embeddings API (requires `OPENAI_API_KEY`)
 3. **mock**: Deterministic hash-based vectors for offline testing
 4. **none**: Disables embeddings (pure lexical scoring)
 
+**Architecture (v0.4.3):**
+- Desktop app: Direct fastembed integration
+- CLI: Shells out to `forest-embed` helper binary
+- Both use identical model → identical search results
+
 Embeddings are computed during capture/edit and stored as JSON arrays in the `embedding` column.
 
 ### Text Processing (src/lib/text.ts)
@@ -568,7 +584,7 @@ FOREST_EMBED_PROVIDER=openai OPENAI_API_KEY=... forest capture --stdin < test.tx
   - `::` - Dual-stack mode, listens on both IPv4 and IPv6
   - `0.0.0.0` - IPv4 only
   - `localhost` - Localhost only (may prefer IPv4 or IPv6 depending on OS)
-- `FOREST_DB_PATH` - Database file path (default: `forest.db` in cwd)
+- `FOREST_DB_PATH` - Database file path (default: see Database Location above)
 - `FOREST_EMBED_PROVIDER` - Embedding provider (default: `local`)
 
 **Network connectivity:**

forest-desktop/CLI-INTEGRATION.md

@@ -24,6 +24,10 @@ Forest.app/
 The CLI is compiled into a standalone executable using `bun build --compile`:
 
 ```bash
+# First, embed sql-wasm.wasm as base64 for standalone distribution
+bun run scripts/embed-wasm.ts
+
+# Then compile to standalone binary
 bun build --compile --minify --sourcemap \
   --outfile forest-desktop/src-tauri/binaries/forest-aarch64-apple-darwin \
   src/index.ts
@@ -32,8 +36,25 @@ bun build --compile --minify --sourcemap \
 This creates a single 65MB binary with:
 - Bun runtime embedded
 - All dependencies bundled
+- **sql.js WASM file embedded as base64** (extracted to temp file at runtime)
+- **forest-embed helper bundled** (Rust binary for semantic embeddings)
 - No need for external `node_modules`
 
+**WASM Embedding Fix (v0.4.1):**
+Previous versions had hardcoded development paths to `sql-wasm.wasm`, causing ENOENT errors in distributed binaries. The fix:
+1. `scripts/embed-wasm.ts` converts WASM to base64 string
+2. `src/lib/db.ts` extracts to temp file at runtime (`/tmp/forest-sql-wasm-<pid>.wasm`)
+3. Temp file is cleaned up on process exit
+4. Falls back to development path when running from source
+
+**Embedding Parity (v0.4.3):**
+CLI and desktop app now use the same embedding engine for true feature parity:
+1. `forest-embed` Rust binary provides semantic embeddings to CLI
+2. Uses `fastembed` crate (same as desktop app)
+3. Both produce identical 384-dim embeddings (all-MiniLM-L6-v2)
+4. CLI shells out to helper binary, falls back to mock if not found
+5. Binary bundled alongside CLI in distribution
+
 ### 2. Tauri Bundle Configuration (`tauri.conf.json`)
 
 ```json
@@ -56,7 +77,25 @@ Provides Tauri commands for:
 - `check_cli_in_path()` - Check if CLI is already in PATH
 - `auto_install_cli_path(shell)` - Auto-add to shell RC files
 
-### 4. Frontend Integration
+### 4. Shared Database Location (v0.4.2)
+
+Both the desktop app and CLI use the same database location for seamless integration:
+
+**Platform-specific locations:**
+- **macOS**: `~/Library/Application Support/com.ettio.forest.desktop/forest.db`
+- **Linux**: `~/.local/share/com.ettio.forest.desktop/forest.db`
+- **Windows**: `%APPDATA%\com.ettio.forest.desktop\forest.db`
+
+**Override:**
+Set `FOREST_DB_PATH` environment variable to use a custom location.
+
+This ensures:
+- ✅ CLI and desktop app always share the same data
+- ✅ Predictable, persistent storage location
+- ✅ Data survives app updates and reinstalls
+- ✅ No confusion about "where is my database"
+
+### 5. Frontend Integration
 
 **Command Palette** (`src/components/CommandPalette.tsx`):
 - Type `/cli-install` to open installation dialog