Skip to content

Setting Up Generative AI

Zenith Filer includes 8 types of AI-powered assistance features. Configure the API provider and key in Control Deck > AI before use.

AI Provider Settings

  • Location: Control Deck > AI
  • Supported providers: Claude / OpenAI / Azure OpenAI / Gemini / Ollama / Claude Code (CLI)
  • API key management: Encrypted with DPAPI and stored locally only (never sent externally). Use "Clear" to remove saved keys
  • Test connection: Verify API key validity with the "Test Connection" button in the API key section. On failure, displays HTTP status code-specific guidance (invalid key, insufficient permissions, rate limit, etc.)
  • Model selection: Choose from a dropdown per provider (defaults used when empty)
  • Model validation: Use the "Validate Model" button in the model section to verify whether the selected model is actually available. Catches model name typos and access permission issues before use
  • Response wait time: Set on the Overview tab. The default, "Automatic", picks the wait from what you are talking to (30 seconds for cloud providers, 2 minutes for reasoning models, 4 minutes for Claude Code, 5 minutes for locally run models such as Ollama). A model running on your own PC can spend several minutes just loading itself the first time, so choose 5 or 10 minutes if replies are being cut off
  • Automatic API format handling for OpenAI / Azure OpenAI: Older models like GPT-4.1 require the Chat Completions API, while newer-generation models like the GPT-5 series require the Responses API. The correct format is inferred automatically from the model name, and if that guess turns out wrong, an error response is detected and the request is automatically retried against the other format — so you can just pick a model without worrying about which API it needs
  • AI ON/OFF: Toggle AI features via the Sparkles icon in the status bar. Long-press to open settings directly
  • Ollama (local LLM) provider: Supports the Ollama provider for local execution without an API key. Features one-click automatic installation, recommended model download management, and automatic server start/stop. When Ollama is not installed, the Ollama tab displays a step-by-step setup guide:
  • Install Ollama — One-click automatic download and installation
  • Start the server — Start the server after installation completes
  • Download a model — Select and download from recommended models

Each step updates its display based on completion status. Once the server is running, the guide disappears and the normal settings screen (endpoint, model selection, recommended model list) is shown - Claude Code (CLI) provider: Runs your locally installed Claude Code in headless mode — no API key needed. If you are logged in with a Claude subscription (Pro/Max), all AI features work without configuring a key - Prerequisites: Claude Code CLI is installed, and you have logged in by running claude in a terminal and using /login - Installation detection: The Claude Code tab shows the detection status and version. The CLI is auto-detected from the default install location (%USERPROFILE%\.local\bin\claude.exe) and PATH; an executable path override is available if needed - Model selection: Choose from default (the CLI's own default model) / sonnet / opus / haiku / fable - Limitations: The MaxTokens/Temperature settings are not applied with this provider. Responses have a few seconds of startup overhead due to the CLI launch, and behavior may change when Claude Code itself is updated - 7-tab AI settings layout: AI settings are divided into 7 sub-tabs: Overview, Ollama, Gemini, Azure OpenAI, OpenAI, Claude, and Claude Code. The Overview tab provides per-feature toggles for automatic AI functions (greeting, rename suggestions, epithet generation, backstory generation, and backup summaries)