Zackriya-Solutions/meetily · error

Failed to create model directory: {}

Error message

Failed to create model directory: {}

What it means

Returned by download_model_detailed when fs::create_dir_all fails on the target model directory (models_dir/<name>, e.g. under ~/Library/Application Support/Meetily/models or %APPDATA%\Meetily\models). The active_downloads entry is removed before the error is returned, so the failed attempt does not block retries.

Source

Thrown at frontend/src-tauri/src/parakeet_engine/parakeet_engine.rs:624

                "nemo128.onnx",
                "vocab.txt",
            ],
            QuantizationType::FP32 => vec![
                "encoder-model.onnx",
                "decoder_joint-model.onnx",
                "nemo128.onnx",
                "vocab.txt",
            ],
        };

        // Create model directory
        let model_dir = &model_info.path;
        if !model_dir.exists() {
            if let Err(e) = fs::create_dir_all(model_dir).await {
                // Remove from active downloads on error
                let mut active = self.active_downloads.write().await;
                active.remove(model_name);
                return Err(anyhow!("Failed to create model directory: {}", e));
            }
        }

        // Clean up incomplete downloads before starting
        log::info!("Checking for incomplete model files to clean up...");
        if let Err(e) = self.clean_incomplete_model_directory(model_dir).await {
            log::warn!("Failed to clean incomplete model directory: {}", e);
            // Continue anyway - we'll handle errors during download
        }

        // Optimized HTTP client for large file downloads
        let client = reqwest::Client::builder()
            .tcp_nodelay(true)              // Disable Nagle's algorithm for better streaming
            .pool_max_idle_per_host(1)      // Keep connection alive
            .timeout(Duration::from_secs(3600))  // 1 hour timeout for large files
            .connect_timeout(Duration::from_secs(30))
            .build()
            .map_err(|e| anyhow!("Failed to create HTTP client: {}", e))?;

View on GitHub (pinned to 0281737d87)

Solutions

1. Check free disk space - the v3 int8 model needs roughly 670 MB plus buffer
2. Verify write permission on the models dir (shown by parakeet_get_models_directory / open_parakeet_models_folder) and fix ownership/permissions
3. If policy or sandboxing blocks the default location, point the app at a writable custom models directory instead of the default
4. Retry the download after fixing - the failed attempt cleaned up its active-download entry

Example fix

// before
engine.download_model(name, None).await?;

// after - preflight the models dir before downloading
let dir = engine.get_models_directory().await;
let probe = dir.join(".write_test");
tokio::fs::write(&probe, b"x").await
    .map_err(|e| anyhow!("models dir {} not writable: {e}", dir.display()))?;
let _ = tokio::fs::remove_file(&probe).await;
engine.download_model(name, None).await?;

// Preflight: writable models dir and enough free space (~700 MB per model)
let dir = engine.get_models_directory().await;
tokio::fs::create_dir_all(&dir).await?; // same op the engine will do
let probe = dir.join(".probe");
tokio::fs::write(&probe, b"x").await?;
let _ = tokio::fs::remove_file(&probe).await;
// check free space via the appropriate platform API before offering the download

match engine.download_model(name, None).await {
    Err(e) if e.to_string().contains("Failed to create model directory") => {
        // surface a dedicated 'check disk space / permissions' message with the dir path;
        // the engine already cleaned its active-download entry, so a retry is safe after fixing
        return Err(e);
    }
    other => other?,
}

When it happens

Trigger: The models directory or its parent is read-only or owned by another user; the disk is full; the path is invalid on the platform (restored backup with wrong permissions, sandboxed app container denying writes); corporate policy blocks writes to Application Support/AppData.

Common situations: Disk full before a ~670 MB download starts; permissions broken after migrating user accounts or restoring from backup; macOS app sandbox or managed-device policy denying the Application Support path; OneDrive/Folder-Redirection on Windows making AppData unavailable.

Related errors

• Failed to create file: {}
• Failed to delete incomplete file {}: {}
• Failed to open file for resume {}: {}
• Failed to write chunk to file: {}
• Failed to flush file: {}

AI-assisted analysis of Zackriya-Solutions/meetily@0281737d87 (2026-08-16). Data as JSON: /api/errors/bebe0e207c01ec54. Report an issue: GitHub.