Installation

Getting Ishvana running on a new machine is a ten-minute job if everything goes right, and maybe twenty minutes if Windows SmartScreen or your antivirus decides to be loud about the installer. The instructions below cover the happy path, the rough edges you’re likely to hit, and the model provider setup that you’ll want to do once the app is installed — because Ishvana without a model provider configured still works, but you’re skipping roughly half of what you paid for. Read all the way through before you start. None of the steps are surprising, but getting them in the right order saves time.

System requirements

• OS. Windows 10 or Windows 11, 64-bit. Windows only
• CPU. A 6-core / 12-thread processor or better. Consumer-grade chips from the last five years are fine; anything older will run Ishvana but might feel sluggish during ML pipelines and local LLM inference.
• RAM. 16 GB minimum. 32 GB if you want to run local LLMs comfortably. 8 GB technically works but you’ll feel the pinch on anything heavy.
• GPU. An Nvidia GPU with 8+ GB VRAM is required for local inference via Ollama. It’s optional if you’re using cloud providers (Anthropic, OpenRouter) instead. GPU optional
• Disk. SSD recommended. Install size is roughly 500 MB plus whatever your projects grow to over time.

The minimum target is a mid-range consumer PC. Ishvana runs fine on older or lower-spec hardware too, but the hardware tier detection in Settings → Performance will drop the app into conservative defaults that keep things responsive at the cost of some features running slower.

Download and install

Three steps, in order:

1. Buy the license. Head to ishvana.com and purchase. After payment, you’ll get a download link and a license key via email. The license key is a long string — copy it somewhere you won’t lose it.
2. Run the installer. The downloaded file is a standard Windows installer (NSIS). Double-click, pick an install location, click through the prompts. The installer is code-signed, so it should work without hassles — but see the troubleshooting section below if Windows SmartScreen gets in the way.
3. Launch Ishvana from the Start Menu or the desktop shortcut.

License activation

On first launch, Ishvana asks for your license key. Paste it into the activation screen and click activate. The app validates the key against the licensing server, registers your machine, and downloads a signed license file that makes subsequent launches work offline — no network required after the first activation.

A few things worth knowing:

• Your license activates on up to three machines by default. Desktop, laptop, and one extra device is usually enough for most authors. If you need more, contact support.
• If you’re moving to a new computer, deactivate on the old machine first via Settings → Licensing. Deactivation frees up a slot; activation on the new machine takes it.
• Lost your key? Check the email you got after purchase. If that’s gone too, contact support with your purchase confirmation.

Full details on how the licensing system actually works — including what happens if your network is down, how offline verification works, and how to deactivate a machine remotely if it’s been lost or destroyed — are in the Licensing settings page.

Auto-updates

Ishvana checks for updates on every launch. When one is available, you’ll see a notification with the release notes. Click to install — the update downloads in the background and applies on the next restart.

Updates are delivered within your purchased major version. Patches and minor feature additions come free; a future Ishvana 6.0 might be a paid upgrade with an existing-customer discount, but that’s a future concern, not something to worry about when you’re installing today.

Where your data lives

Ishvana stores everything under your Windows AppData folder at C:\Users\<YourName>\AppData\Roaming\Ishvana\.

• DirectoryIshvana/

  • Directorydata/

    • Directorydocuments/ (.docx manuscript files)

      • …
    • lore.db (SQLite Legendry database)
    • ishvana.db (SQLite main database)
    • Directorychromadb/ (vector embeddings)

      • …
    • Directoryattachments/ (images and assets)

      • …
    • Directorybackups/ (auto-generated backups)

      • …
    • Directoryconfig/ (settings)

      • …
    • Directorylogs/ (error logs and crash dumps)

      • …
  • (electron app state)

This is the directory you want to back up regularly. Your manuscripts, your Legendry, your agent conversation history, your research bookmarks — all of it lives here. For the full breakdown of what every folder contains and why, see Data Storage.

Back up your data

Ishvana does its own auto-backups into data/backups/, but those live in the same directory as your main data. If the whole data folder gets corrupted or deleted, the auto-backups go with it. You want a second backup somewhere — another drive, Backblaze, Dropbox, whatever works for you — so that a bad day doesn’t turn into a lost novel.

Model provider setup

Ishvana’s agents need a language model provider to work. You pick one (or several) and configure them in Settings → Models after installation. Three providers are supported:

Ollama (local)

Runs models directly on your machine. Free — no API costs after the hardware is paid for. Private — your data never leaves your computer. Offline — works without an internet connection. Slower than cloud providers for most models, and needs a capable GPU.

1. Install Ollama separately from ollama.com.
2. Pull a model. ollama pull llama3 or ollama pull mistral or whatever you’d like to try.
3. In Ishvana’s Settings, set the Ollama base URL (default is http://localhost:11434).
4. Ishvana’s model management interface lets you pull and delete models directly from the app after initial setup.

Best for: writers who want zero ongoing cost, who value privacy, or who work offline frequently.

Anthropic (direct)

Direct access to Claude models via API key. The best models available for structured output and long-context work — Claude Opus in particular has a measurable edge on the kind of consistency checking and creative reasoning Ishvana’s agents do.

1. Sign up at console.anthropic.com.
2. Add API credits and generate an API key. Keys start with sk-ant-.
3. Paste the key into Ishvana’s Settings under the Anthropic provider section.

Best for: writers who want the highest-quality output and are willing to pay per-token. Expect to spend something like $5-20 per heavy writing session depending on how much you use the deeper reasoning features.

OpenRouter (aggregated)

A routing service that gives you access to many models from many providers through a single API key. Good for experimentation — you can try GPT-4, Claude, Gemini, open-source models, and niche fine-tunes all from one credential.

1. Sign up at openrouter.ai.
2. Add credits and generate an API key.
3. Paste the key into Ishvana’s Settings under the OpenRouter provider section.
4. Browse and assign models from Ishvana’s model picker.

Best for: writers still figuring out which model works best for them, or writers who want to compare multiple models without managing separate provider credentials.

You can mix providers

Nothing stops you from running Ollama for quick tasks (entity extraction, tag generation, inline transforms) and Anthropic for deep reasoning (Lorekeeper checks, Creative Studio debates, long-manuscript analysis). Ishvana’s smart router handles the routing automatically once you configure the tier assignments. Most serious users end up with a multi-provider setup within a month.

For the conceptual background on how Ishvana routes between providers and why, see Providers & Models.

Troubleshooting

The rough edges that most often come up on first launch:

Windows SmartScreen warning

The installer is code-signed through Azure Trusted Signing, but SmartScreen can still flag it as “unrecognized” on less-common software — especially for authors downloading it for the first time. If you see the warning, click More info, then Run anyway. This is standard SmartScreen behavior for new software, not a sign that anything’s wrong.

Ishvana won’t start after install

Most of the time this is antivirus. Ishvana runs a Python backend as a child process on port 37737, and some security tools don’t like that pattern. Add an exception for the Ishvana install directory in your antivirus — Windows Defender, Malwarebytes, Norton, whatever you use. If that doesn’t help, check the logs in %APPDATA%\Ishvana\data\logs\ for the actual error.

Antivirus false positives

Some antivirus tools flag Ishvana’s backend as suspicious because it’s a PyInstaller-bundled Python executable, and PyInstaller binaries trigger heuristics in certain scanners. This is a false positive. Ishvana is code-signed and safe; the scanners are being overly eager. Add an exception and move on.

License key issues

Paste errors are the most common cause. Check for extra spaces at the start or end of the key when you paste it. If you’re still getting rejections after a clean paste, contact support with your purchase confirmation email and they’ll sort it out.

None of these are common, and most users get through installation without hitting any of them. But if you’re the unlucky one who hits all four on the same day, at least the fixes are quick.

What’s next

Your First Project Create a project, add your first lore entry, write a scene, and talk to an agent — all in one sitting.

Settings Configure model providers, editor preferences, performance tuning, and everything else you can tune.

Licensing Full details on how activation, deactivation, and license management work.