Resolve setup issues.
Get Sparrow running.

Operating System

Use Windows 11, 64-bit, on an Intel or AMD x64 PC. Windows on ARM is not covered by this installer. Install available Windows updates and restart if a reboot is pending before running setup again.

Check your Windows version and architecture

Open Settings → System → About and review Windows specifications and System type. You can also run these commands in PowerShell:

Get-CimInstance Win32_OperatingSystem |
    Select-Object Caption, Version, OSArchitecture

[System.Runtime.InteropServices.RuntimeInformation]::OSArchitecture

The operating system architecture should be X64. Run the Sparrow installer as an administrator. If Windows blocks the installer, verify its source and review the exact message before proceeding.

Architecture

Sparrow Self-Hosted requires an x64 operating system and an Intel or AMD x64 processor. A 32-bit Windows installation or an ARM64 computer is not covered by this installer.

Check your system architecture

Open Settings → System → About and check System type. It should show a 64-bit operating system and an x64-based processor. In PowerShell, run:

[System.Runtime.InteropServices.RuntimeInformation]::OSArchitecture
[System.Runtime.InteropServices.RuntimeInformation]::ProcessArchitecture

The operating system result should be X64. The process result describes the PowerShell session itself; an X86 process can run on x64 Windows. Use 64-bit PowerShell when checking the native environment.

Memory

Plan for at least 16 GB of system RAM; 32 GB or more is recommended for additional processing headroom. Available memory also depends on other applications and background services.

Check installed and available memory

Open Task Manager → Performance → Memory, or run these PowerShell commands:

Get-CimInstance Win32_ComputerSystem |
    Select-Object @{Name='InstalledRAMGB';Expression={[math]::Round($_.TotalPhysicalMemory / 1GB, 1)}}

Get-CimInstance Win32_OperatingSystem |
    Select-Object @{Name='AvailableRAMGB';Expression={[math]::Round($_.FreePhysicalMemory / 1MB, 1)}}

If memory usage is high

Close unnecessary applications and inspect memory usage in Task Manager. If Sparrow repeatedly runs out of memory, review the camera workload and model settings or increase installed RAM. Keep the Windows paging file system-managed unless your administrator has a specific configuration.

Storage

Install Sparrow on an SSD. As a planning baseline, allow at least 20 GB free for software, AI models, and initial data; 50 GB or more is recommended. Event images, logs, and retention require additional capacity.

Check free disk space

Get-PSDrive -PSProvider FileSystem |
    Select-Object Name,
        @{Name='FreeGB';Expression={[math]::Round($_.Free / 1GB, 1)}},
        @{Name='UsedGB';Expression={[math]::Round($_.Used / 1GB, 1)}}

Check both the installation drive and the Windows temporary-files drive. Downloads and extraction may need space on each. Confirm that the service account can read the model files and write to the configured data folders.

If an AI model download is incomplete

Free sufficient disk space, stop the affected Sparrow service, and retry the download through setup. If your release provides an expected SHA-256 checksum, compare it with the downloaded file:

# Replace this example path with the actual model file.
Get-FileHash -Algorithm SHA256 -Path 'C:\path\to\model.gguf'

NVIDIA GPU

Sparrow’s NVIDIA CUDA build requires a compatible NVIDIA GPU and driver. The current hardware planning baseline is an NVIDIA RTX GPU with 6 GB dedicated VRAM; an RTX 4060 with 8 GB VRAM or a suitable higher-capacity GPU is recommended. Validate your model and camera workload before choosing hardware.

Check the GPU and driver

Open Command Prompt or PowerShell and run:

nvidia-smi

nvidia-smi --query-gpu=name,driver_version,memory.total --format=csv

If the command fails, check Device Manager and the NVIDIA driver installation. Install a driver compatible with the CUDA runtime supplied with your Sparrow release, then reboot before testing again.

If CUDA initialization fails or GPU memory is exhausted

• Confirm that Windows detects the intended NVIDIA GPU and that its driver loads successfully.
• Close other GPU-heavy applications and inspect available GPU memory with nvidia-smi.
• Use the runtime libraries supplied for the same Sparrow release. Repair or rerun setup if a CUDA DLL is missing.
• Restart RTVCP-Overlord after resolving the driver or runtime issue, then check the application error again.

The CUDA version shown by nvidia-smi describes driver capability; it does not confirm that the CUDA development toolkit is installed. End users normally need the release’s runtime dependencies and compatible driver, rather than the compiler toolkit.

Port 5432 availability

If another PostgreSQL installation is already using TCP port 5432, configure that installation for Sparrow. Create a database named rtvcp with a login role named rtvcp using the password rtvcp. Make sure PostgreSQL is available only for localhost connections.

Check the port in PowerShell

Get-NetTCPConnection -LocalPort 5432 -State Listen -ErrorAction SilentlyContinue
Test-NetConnection -ComputerName 127.0.0.1 -Port 5432

A successful TCP test confirms a reachable listener. Verify that the listener belongs to PostgreSQL and that the database credentials work.

Create the login and database

Connect to the existing installation as a PostgreSQL administrator using pgAdmin or psql. Run these statements separately, outside a transaction, only if the role and database do not already exist:

CREATE ROLE rtvcp WITH LOGIN PASSWORD 'rtvcp';
CREATE DATABASE rtvcp OWNER rtvcp;

If the role or database already exists, check its password, ownership, and permissions instead of creating duplicates. Do not grant the Sparrow role superuser privileges.

Restrict access to this computer

In the active PostgreSQL postgresql.conf, use:

listen_addresses = 'localhost'
port = 5432

In pg_hba.conf, allow password-authenticated loopback connections for Sparrow:

host    rtvcp    rtvcp    127.0.0.1/32    scram-sha-256
host    rtvcp    rtvcp    ::1/128         scram-sha-256

Review existing access rules and restart the PostgreSQL service after changing its listening address. If other applications require remote access to this PostgreSQL installation, contact your administrator before changing its configuration.

Port 6379 availability

If another Redis installation is already using TCP port 6379, make sure it is running and available for local connections from Sparrow. Two services cannot listen on the same address and port at the same time.

Check the port in PowerShell

Get-NetTCPConnection -LocalPort 6379 -State Listen -ErrorAction SilentlyContinue
Test-NetConnection -ComputerName 127.0.0.1 -Port 6379

Verify that Redis responds

If redis-cli is installed, run:

redis-cli -h 127.0.0.1 -p 6379 PING

The expected response is PONG. An authentication error means Redis is reachable but requires credentials; configure matching credentials in Sparrow if supported. Do not disable existing authentication just to bypass the error.

Allow loopback connections

For a standard Redis configuration, bind to loopback and keep protected mode enabled:

bind 127.0.0.1 -::1
protected-mode yes
port 6379

Restart Redis after changing its configuration. Configuration options differ for Redis-compatible services; use that service’s equivalent loopback binding settings. If Redis is shared with another application, confirm database selection and key isolation before reusing it. Do not clear an existing cache as a troubleshooting step.

Start, stop, or restart RTVCP services

Manage RTVCP-Overlord and RTVCP-Update from Windows Services or PowerShell. Open PowerShell with Run as administrator before running these commands.

Check service status

Get-Service -Name 'RTVCP-Overlord', 'RTVCP-Update' |
    Format-Table Name, DisplayName, Status -AutoSize

RTVCP-Overlord

Use the command for the action you need. Stopping or restarting this service may interrupt Sparrow operation.

# Start
Start-Service -Name 'RTVCP-Overlord'

# Stop
Stop-Service -Name 'RTVCP-Overlord'

# Restart (also starts it if currently stopped)
Restart-Service -Name 'RTVCP-Overlord'

RTVCP-Update

Use the command for the action you need. Avoid stopping or restarting the update service while an update is being applied.

# Start
Start-Service -Name 'RTVCP-Update'

# Stop
Stop-Service -Name 'RTVCP-Update'

# Restart (also starts it if currently stopped)
Restart-Service -Name 'RTVCP-Update'

Use the Windows Services application

1. Press Win + R, enter services.msc, and press Enter. Accept the administrator prompt if shown.
2. Locate RTVCP-Overlord or RTVCP-Update. If the display name differs, open its properties and check the Service name.
3. Right-click the service and select Start, Stop, or Restart, as available.
4. Refresh the list and confirm the resulting status. A stopped service may show a blank Status column.

If a service does not start

Record the error, check PostgreSQL and Redis connectivity using the sections above, and review Event Viewer → Windows Logs → Application and System for related errors. If Windows cannot find the service name, confirm that setup completed and check its registered name in Services.