Setup wizard

The setup wizard is an interactive terminal interface for configuring LLM providers, validating API keys, and setting default model routing.

Launch the wizard from your terminal:

bash

seepient setup

Seepient Setup Wizard — Select an action:

1. Add or configure an LLM provider
2. Configure Purpose × Tier model routing
3. Test provider connections
4. Configure consent mode & sandbox
5. Exit

1. Provider configuration

When you select Add or configure an LLM provider, Seepient lists supported providers discovered from the upstream catalog:

OpenAI
Anthropic Claude
Google Gemini
Zhipu GLM
DeepSeek
OpenRouter
Local endpoints (Ollama, vLLM, LM Studio)
Custom OpenAI-compatible endpoints

Credential storage modes

The wizard supports three ways to store credentials:

Paste the key (default): The key is stored through provider management (encrypted credential file, or the OS keychain when available) — never read back from the environment at runtime.
Keyless / local: Configures local models (like Ollama on http://127.0.0.1:11434) that require no authentication tokens.
Sign in with provider: OAuth sign-in for subscription providers (shown when the provider supports an OAuth flow).

2. Live model catalog discovery

After authenticating with a provider, the wizard queries the provider's /models endpoint to discover available models in real time.

For each model, Seepient displays:

Context window size (for example, 200k tokens)
Input, output, and cache token pricing
Reasoning token support
Vision and tool capability flags

3. Purpose × Tier assignment

The wizard guides you through assigning models to operational roles:

Text (Standard): The default companion model for general tasks.
Commit (Coding): The model responsible for analyzing code, formatting patches, and writing files.
Complex Reasoning: The model invoked when a task requires deep planning or architectural decomposition.
Media: The backend used when generating images or visual mockups.

4. Connection verification

Before saving changes to disk, the wizard sends a minimal test ping to each configured provider. If the request fails (due to an invalid API key, network timeout, or billing restriction), the wizard highlights the error details and prompts you to correct the credentials.

Settings are written atomically to ~/.seepient/setting.json. Existing comments, workspace-specific overrides, and unrelated settings are preserved.

Setup wizard ​

1. Provider configuration ​

Credential storage modes ​

2. Live model catalog discovery ​