Skip to content

Settings Walkthrough

The settings modal has six tabs and a couple dozen individual knobs, and a fair number of users open it for the first time on day one, feel overwhelmed, close it, and never come back. That’s fine — most of the defaults are tuned to just work, and you genuinely don’t need to touch the majority of settings to use Ishvana day to day. But a few settings matter a lot for getting Ishvana actually working the way you want it to, and this page walks through which ones those are so you don’t spend an hour in the modal when you really only need to touch four things. Everything below is the short version. For the full reference on every setting, click through to the dedicated Settings pages under the Settings section.

Open Settings from the gear icon in the top bar. Six tabs: General, Editor, Models, Email Accounts, Performance & Hardware, and Licensing.

The four knobs you actually need to set before Ishvana feels usable:

  1. Model provider credentials — otherwise the agents can’t do anything. This is the most important setup step. See Models below.
  2. License activation — required for anything other than the trial to work. See Licensing below.
  3. Theme — dark or light, purely a preference. Takes about three seconds.
  4. Auto-save interval — the default is 2 seconds, which works for most people. Only adjust if you have a specific reason.

Everything else can wait. The Email tab doesn’t matter until you decide to use Ishvana’s email features. Performance tuning doesn’t matter until something feels slow. Per-agent model overrides don’t matter until you’re running cost analytics and want to pin a specific model to a specific task.

Theme, notifications, analytics toggle, and a few other app-wide preferences. Most users set this once on day one and never open it again.

  • Theme. Dark or Light. The default is Dark because most writing sessions happen in low-light conditions. Switch if you prefer Light.
  • Notifications. All, Important, or None. The default is Important, which surfaces errors and major events but stays quiet about routine things like save confirmations.
  • Analytics. Track writing sessions and goals. Default on. Disable if you don’t want the writing stats panel to collect data — though the stats are local-only, not uploaded anywhere, so privacy isn’t a real concern.

Full reference: Settings → General.

How the writing surface behaves. The auto-save interval is the one people actually tweak; the rest are set-and-forget.

  • Auto-save. On by default. Saves every 2 seconds incrementally. Leave it on.
  • Auto-save interval. 0.5 to 10 seconds. Default 2. Only adjust if you have a specific reason.
  • Spell check. Uses the OS-native spell checker. Default on. Add custom words to your personal dictionary via right-click.
  • Content width. 600 to 1400 pixels. Default 760 — the width that produces comfortable 65-85 character lines. Widen if you prefer more text per line; narrow if you want a single-column printed-page feel.

Full reference: Settings → Editor.

The most technically involved tab. This is where you set up the model provider credentials that the agent pipeline depends on. Without at least one provider configured, Hawken can’t run edit passes or sketch anything, Lorekeeper can’t run consistency checks, Lagan can’t do research, and most of what you bought Ishvana for isn’t working.

Three providers are supported:

  • Ollama. Runs models locally on your machine. Free once hardware is paid for. Private. Offline. Requires a capable GPU. Set the base URL (default http://localhost:11434).
  • Anthropic. Direct Claude API access. Highest quality. Per-token billing. Enter your API key starting with sk-ant-.
  • OpenRouter. A routing service with access to many models. Best for experimentation. Per-token billing with a small margin.

Pick one (or several) and enter the credentials. After saving, the model picker populates with the models available from your configured providers, and you can assign specific models to the two tiers (Quick and Reasoning) and to per-agent overrides.

The tab also includes the context compression settings (auto-compression, history snip, max output recovery) and budget limits for LLM spending. Most users should leave all of those at their defaults for the first week of use and adjust only if something specific comes up.

Full reference: Settings → Models. Conceptual background: Engine → Providers & Models.

Only matters if you plan to use Ishvana’s Email module for author correspondence. If you’re handling email in Gmail or another client and you don’t want Ishvana to touch it, skip this tab entirely — it’s optional.

If you are going to use it: add each of your email accounts with SMTP and IMAP credentials, test the connection, done. Most providers (Gmail, Fastmail, Outlook) require an app password rather than your regular account password.

Full reference: Settings → Email Accounts.

Hardware tier detection and performance tuning. You probably won’t visit this tab unless something feels slow. On first launch, Ishvana detects your machine’s specs and picks reasonable defaults (high-end / balanced / standard / basic tier). The defaults work for most users.

What’s here if you do visit:

  • Detected hardware tier (read-only). The classification Ishvana picked for your machine.
  • GPU probe. Test whether your GPU is detected and usable. Useful after driver updates.
  • Alert thresholds. When to warn you about CPU or memory pressure.
  • Cache TTL and size. How long cached data stays valid and how much memory the cache can use.
  • Undo depth. How many undo steps the editor keeps in memory per document.

Full reference: Settings → Performance & Hardware.

License activation, machine management, and trial status. You’ll visit this tab once to activate your license on first launch and never again unless you’re moving to a new machine or something goes wrong.

  • Enter a license key to activate on this machine.
  • Start a 14-day trial by entering your email if you don’t have a license yet.
  • View activation status — license type, expiration, machines used vs. limit.
  • Deactivate this machine to free up a slot when you’re moving to a new computer.

Full reference: Settings → Licensing.