Chapter 18: Settings
The Settings screen allows detailed customization of ThinkNavi’s behavior. Access by clicking “Settings” in the sidebar.
18.1 Language Settings
Operation:
- Click the “English” or “Japanese” button at the top
- The entire interface language switches immediately
Notes:
- UI text (menus, buttons, labels) switches
- AI response language does not change. AI responds in the language of the question
- Settings are saved in the browser and persist across logins
18.2 Image Generation Settings
Configure the provider and model for AI image generation.
Provider Selection:
- Google Gemini: Google’s AI image generation. Gemini API key required
- OpenAI: DALL-E based image generation. OpenAI API key required
Custom Image Model:
Enter a specific model ID in the text input.
Examples:
gemini-2.0-flash-exp— Gemini experimental versiongpt-image-1— OpenAI’s latest image generation modeldall-e-3— DALL-E 3
Leave blank to use the selected provider’s default model.
18.3 Thinking Mode Settings
Configure the thinking enhancement mode used in the Thinking Support Chat.
| Mode | Description | Recommended Use |
|---|---|---|
| Off | AI responds directly. No thinking techniques applied | Simple questions, information lookup |
| Rule (Recommended) | Automatically selects and applies thinking techniques based on keyword rules | Daily thinking support |
| Observer (Experimental) | Auxiliary LLM analyzes the question and dynamically selects optimal techniques. More advanced but slightly slower responses | Complex problem analysis, research |
Auto-Applied Thinking Techniques:
When Rule or Observer mode is active, the following techniques are automatically applied based on question content:
| Technique | Trigger Examples (Rule Mode) | Effect on AI Response |
|---|---|---|
| Analogy | “like”, “compare to”, “for example” | Explains using concrete examples from other domains |
| Fermi Estimation | “how many”, “approximately”, “estimate” | Derives estimates through step-by-step reasoning |
| Debate | “should we”, “pros and cons”, “advantages/disadvantages” | Develops arguments from both for and against |
| Multiple Perspectives | “other viewpoints”, “different angle”, “from X’s perspective” | Analyzes from multiple positions and viewpoints |
| Role Model | “what would X do”, “from X’s standpoint” | Thinks from a specific person or role’s perspective |
| Metacognition | “assumptions”, “organize the issues”, “thinking process” | Analyzes and organizes the structure of thinking itself |
18.4 Voice Settings
Select the voice used for Text-to-Speech.
| Voice | Characteristics |
|---|---|
| Alloy | Neutral, calm voice |
| Echo | Slightly lower tone |
| Fable | Warm voice |
| Onyx | Deep, masculine voice |
| Nova (default) | Bright, natural voice |
| Shimmer | Soft, gentle voice |
Selected voice applies to chat playback and auto-response after voice input.
18.5 AI Model Selection
Configure the AI model used in chat.
Purpose-Based Auto-Selection
Select a purpose from the “Purpose” dropdown to auto-select the optimal, most cost-efficient model.
| Purpose | Auto-Selected Model Tendency | Suited For |
|---|---|---|
| ——— | —————————– | ———– |
| Latest Info | Real-time search models (Perplexity, etc.) | News, latest tech info |
| Logical Reasoning | High reasoning models (GPT-4o, etc.) | Analysis, calculation, programming |
| Empathetic | High dialogue models (Claude, etc.) | Consultation, brainstorming |
| Low Cost | Lightweight models (GPT-4o-mini, etc.) | Simple questions, bulk processing |
Manual Model Selection
Select a specific model from the “Current Model” dropdown. Models are grouped by provider.
Available model examples:
- OpenAI: gpt-4o, gpt-4o-mini, o1, o1-mini, o3-mini
- Anthropic: claude-3-5-sonnet, claude-3-5-haiku (when API key is set)
- Google: gemini-2.0-flash, gemini-1.5-pro (when API key is set)
- xAI: grok-2 (when API key is set)
- Perplexity: sonar-reasoning (when API key is set)
Model Reset Timer: Manually selected models revert to default after a certain time. A timer is displayed on screen.
Custom Model Input
You can also enter a specific model ID directly in the text input.
Examples:
gpt-4-turbo-2024-04-09(specific version)claude-3-opus-20240229gemini-1.5-pro-002
18.6 API Key Management
Register API keys for external AI providers. Registering an API key adds that provider’s models to the model selection list.
Supported Providers
| Provider | Key Format | Registration Benefits |
|---|---|---|
| OpenAI | String starting with sk- | 50% credit discount + full GPT-4o features |
| Anthropic | String starting with sk-ant- | Claude series models available |
| Google Cloud API key | Gemini series models available | |
| xAI | xAI API key | Grok series models available |
| Perplexity | Perplexity API key | Real-time search models available |
| IEEE Xplore | IEEE Developer API key | IEEE academic paper search available |
API Key Registration Steps
- Expand the “API Keys” section in Settings
- Enter the key in the text input for the desired provider
- Input is automatically saved and validated
- Valid keys add corresponding models to the selection list
- Invalid keys display an error message
Security:
- API keys are encrypted and stored on the server
- Only the first and last few characters are displayed (e.g.,
sk-...xxxx) - To delete a key, clear the input field
OpenAI API Key Discount Details
Registering your own OpenAI API key gives you a 50% discount on ThinkNavi credit consumption.
How the Discount Works:
- Requests are sent directly using your own OpenAI API key, not through ThinkNavi’s server
- Since ThinkNavi’s server costs are eliminated, credit consumption is halved
- OpenAI usage fees are charged directly to your OpenAI account
- The discount applies to all OpenAI operations: chat, embedding, image generation, etc.
Example: A chat message normally costs 1 credit, but with your own key it costs 0.5 credits
18.7 Settings Troubleshooting
| Issue | Cause and Solution |
|---|---|
| API key entered but models don’t appear | Reload the page. Check that the key format is correct (OpenAI starts with sk-) |
| Thinking mode change not reflected | Start a new conversation. Previous settings apply to existing conversations |
| Voice not playing | Check browser audio output settings. Safari has auto-play restrictions |
| Language change doesn’t affect AI responses | UI language and AI response language are independent. AI responds in the question’s language. Ask in English for English answers |
| Custom model gives error | Check that the model ID is accurate. The provider’s API key must be registered |