Skip to main content

​
Overview

PolyChat-AI requires an OpenRouter API key to access 100+ language models. This guide walks you through the initial setup and configuration options.

​
API Key Setup

1

Get your OpenRouter API key

  1. Visit OpenRouter.ai
  2. Create a free account
  3. Navigate to API Keys
  4. Generate a new API key
2

Configure PolyChat-AI

Automatic Configuration (First Launch)On first launch, an onboarding modal will appear prompting you to enter your API key.Manual Configuration
  1. Open Settings using Ctrl/Cmd + K or click the ⚙️ icon
  2. Navigate to the “API” section
  3. Enter your OpenRouter API key (format: sk-or-v1-...)
  4. Click “Save”
Your API key is securely stored in your browser’s localStorage.
3

Verify Connection

After saving your API key, check the “System Information” section in Settings:
  • API Status: Should show “Connected”
  • Available Models: Should display the number of accessible models (100+)

​
Model Selection

​
Default Model

Set your preferred default model for new conversations:
  1. Open Settings (Ctrl/Cmd + K)
  2. Go to “Default Model” section
  3. Use the search bar to filter models by name or provider
  4. Click on a model to select it
Model Information Display:
ColumnDescription
ModelModel name and provider
CtxContext window size (in thousands of tokens)
PriceInput/Output pricing per 1M tokens
✓Currently selected model

​
Keyboard Navigation

When searching for models in the Settings modal:
  • ↑ / ↓ - Navigate through model list
  • Enter - Select focused model
  • Type to search by model name or provider

​
Model Switching Mid-Conversation

You can change models during an active conversation:
  1. Click the model switcher in the header
  2. Select a different model from the dropdown
  3. Your next message will use the newly selected model

​
Multi-Model Chat

Compare responses from up to 3 models simultaneously:
  1. Click the multi-model icon to enable grid view
  2. Select up to 3 different models
  3. Send a message to see side-by-side responses
  4. Each response shows which model generated it

​
System Instructions

​
Custom System Prompts

Define how the AI should behave across all conversations:
  1. Open Settings (Ctrl/Cmd + K)
  2. Navigate to “System Instruction” section
  3. Enter your custom prompt (e.g., “You are a specialist in…”)
  4. This instruction is sent at the beginning of each new conversation
Example System Prompts: 
You are an expert software engineer specializing in TypeScript, React, and modern web development. Provide clear, concise code examples with explanations. Always follow best practices and consider performance implications.

You are a professional technical writer. Create clear, well-structured documentation with proper formatting. Use examples and diagrams when helpful. Write for developers of all skill levels.

You are a data analysis expert. Provide statistical insights, interpret data patterns, and suggest visualization approaches. Explain complex concepts in accessible language.

​
Conversation Tone

Choose the AI’s response style:
ToneDescription
NeutralBalanced, objective responses
FormalProfessional, structured communication
FriendlyWarm, conversational style
ProfessionalBusiness-oriented, polished responses
EnthusiasticEnergetic, positive communication
The selected tone is automatically added to your system instruction.

​
Advanced Settings

​
RAG (Retrieval Augmented Generation)

Status: Experimental Enables intelligent context enhancement using local embeddings:
  • Technology: Uses @xenova/transformers with all-MiniLM-L6-v2 model
  • Privacy: All processing done locally in your browser
  • Functionality: Retrieves semantically relevant conversation history
  • Default: Enabled
To toggle RAG:
  1. Open Settings (Ctrl/Cmd + K)
  2. Navigate to “AI Features” section
  3. Toggle “Context Enhancement (RAG)“

​
Notifications

Configure browser notifications for AI responses:
  1. Open Settings (Ctrl/Cmd + K)
  2. Navigate to “Notifications” section
  3. Toggle “Browser Notifications”
  4. Grant browser permission when prompted
Note: Notifications alert you when a response is ready, useful when working in other tabs.

​
Configuration Storage

All settings are stored locally in your browser:
  • Storage Method: localStorage
  • Persistence: Settings persist across browser sessions
  • Privacy: No cloud synchronization (data stays on your device)
  • Export: Currently no export feature (see roadmap)
Clearing your browser data will reset all settings and conversation history. Consider exporting important conversations before clearing browser storage.

​
Troubleshooting

  1. Verify the key format starts with sk-or-v1-
  2. Check that the key hasn’t expired on OpenRouter.ai
  3. Ensure you have credits or are within free tier limits
  4. Try regenerating a new key on OpenRouter
  1. Verify your API key is correctly configured
  2. Check your internet connection
  3. Refresh the page to reload model list
  4. Check OpenRouter status page for API issues
  1. System prompts only apply to new conversations
  2. Start a new conversation (Ctrl/Cmd + N) to test
  3. Verify the prompt was saved in Settings
  4. Check that the prompt doesn’t exceed model limits

​
Next Steps

Customization

Personalize your interface with themes and accent colors

Keyboard Shortcuts

Master all keyboard shortcuts for efficient workflows

Build docs developers (and LLMs) love

Get started for freeTalk to us