Docs Playground
Open Playground

Playground Guide

The Plexor Labs Playground is an interactive chat interface for testing intelligent LLM routing in real-time. See how requests are routed to different providers, compare response quality, and watch your savings accumulate.

On This Page

What is the Playground?

The Plexor Labs Playground is a web-based chat interface that lets you interact with AI models through Plexor Labs' intelligent routing system. It provides a real-time demonstration of how Plexor Labs optimizes your LLM costs by routing requests to the most appropriate provider.

Key Features

Real-Time Savings Display

Watch your savings accumulate as you chat. Each response shows which provider handled it and how much you saved compared to using the premium model directly.

Multiple Models

Access Claude, GPT-4o, DeepSeek, Gemini, and Mistral through a single interface. Let Plexor Labs choose automatically or force a specific provider.

Conversation History

Your conversations are automatically saved and synced across devices. Continue where you left off or review past interactions.

Getting Started

You can use the Playground in two ways: as a logged-in user with your account credentials, or as a guest using a Plexor Labs API key.

Access the Playground

Navigate to plexor.dev/playground from your browser. If you're logged in, you'll be ready to start chatting immediately.

Authenticate (if needed)

If you're not logged in, click the key icon in the top-right corner to enter your Plexor Labs API key. Get your key from the API Keys page.

Start Chatting

Type your message in the input field at the bottom and press Enter or click the send button. Your first message will create a new conversation automatically.

Pro Tip
Logged-in users get cloud sync for conversations across devices. Guest users with API keys have conversations stored locally in the browser.

Interface Overview

The Playground interface is designed to be intuitive and familiar if you've used ChatGPT or Claude.

Sidebar (Left)

  • New Chat Button - Start a fresh conversation
  • Conversation List - View and switch between past conversations
  • Delete Button - Remove conversations (hover to reveal)
  • User Menu - View account info and sign out

Header (Top)

  • Model Selector - Choose between Auto, Claude, GPT-4o, DeepSeek, Gemini, or Mistral
  • Mode Selector - Switch between Eco, Balanced, and Quality modes
  • Savings Badge - Shows your cumulative savings for the session
  • Theme Toggle - Switch between dark and light modes
  • API Key Button - Configure your API key (for guest users)

Chat Area (Center)

  • Welcome Screen - Shown when no conversation is active
  • Messages - Your conversation with AI responses
  • Provider Badge - Shows which provider handled each response

Input Area (Bottom)

  • Attach Button (+) - Add files to your message
  • Text Input - Type your message (supports multi-line with Shift+Enter)
  • Send Button - Send your message (or press Enter)

Model Selection

The model selector dropdown lets you choose how your requests are routed.

Model Description Best For
Auto (Best Value) Plexor Labs intelligently routes to the optimal provider General use, maximum savings
Claude 3.5 Sonnet Force all requests to Anthropic Claude Complex reasoning, code generation
Claude 3 Opus Force all requests to Claude's most powerful model Critical tasks requiring highest quality
GPT-4o Force all requests to OpenAI GPT-4o Multimodal tasks, creative writing
Gemini 2.0 Flash Force all requests to Google Gemini Fast responses, Google ecosystem
Mistral Large Force all requests to Mistral AI European data residency, multilingual
DeepSeek Chat Force all requests to DeepSeek Cost-effective for simpler tasks
Auto Mode
When set to "Auto (Best Value)", Plexor Labs analyzes each request and routes it to the most cost-effective provider capable of handling it well. Simple questions go to cheaper models, while complex tasks automatically use premium providers.

Routing Modes

The routing mode determines how aggressively Plexor Labs optimizes for cost savings.

Eco Mode

Maximum cost savings. Routes most requests to the cheapest capable providers. Best for learning, exploration, prototyping, and simple tasks where cost is the priority.

  • Typical savings: 70-85%
  • Routes aggressively to DeepSeek and Mistral
  • Only uses premium models for truly complex tasks

Balanced Mode

The default mode. Balances cost savings with response quality. Suitable for most general development and production work.

  • Typical savings: 60-75%
  • Uses Claude Sonnet for moderate complexity
  • Good balance between quality and cost

Quality Mode

Prioritizes response quality. Uses premium models more frequently. Best for critical code, complex architecture decisions, or sensitive work.

  • Typical savings: 30-50%
  • Prefers Claude Sonnet and Opus
  • Still avoids premium models for trivial requests
Example: 100 Request Session
$8.50
Average savings in Balanced mode vs. using Claude directly for all requests

Managing Conversations

The Playground automatically manages your conversation history.

Creating Conversations

A new conversation is created automatically when you send your first message from the welcome screen. The conversation title is generated from your first message.

Switching Conversations

Click any conversation in the sidebar to load it. Your current conversation is automatically saved.

Deleting Conversations

Hover over a conversation in the sidebar and click the trash icon to delete it. You'll be asked to confirm before deletion.

Cloud Sync (Logged-in Users)

If you're signed in, your conversations are automatically synced to the cloud. Access them from any device by logging into your account.

Guest User Storage
If you're using an API key without logging in, conversations are stored in your browser's local storage. They won't sync across devices and will be lost if you clear browser data.

File Attachments

You can attach text files to your messages to provide context or ask questions about code.

Supported File Types

The Playground accepts common text-based files including:

  • Code files: .py, .js, .ts, .jsx, .tsx, .go, .rs, .java, .c, .cpp, .rb, .php, .swift, .kt, .scala
  • Config files: .json, .yaml, .yml, .xml
  • Web files: .html, .css
  • Text files: .txt, .md, .sql, .sh, .bash

File Size Limits

  • Maximum file size: 100KB per file
  • Maximum total attachment size: 500KB per message

How to Attach Files

  1. Click the + button next to the message input
  2. Select one or more files from your computer
  3. Files appear as chips above the input field
  4. Click the X on a chip to remove a file
  5. Type your message and send

Tracking Savings

The Playground shows your real-time savings in the header.

Savings Badge

The green badge in the top-right corner shows your cumulative savings for the current session. This compares what you would have paid using Claude directly versus what you actually paid through Plexor Labs.

Provider Badges

Each AI response includes a small badge showing which provider handled it. Common badges include:

  • DEEPSEEK - Request was routed to DeepSeek (cheaper)
  • CLAUDE-SONNET - Request was routed to Claude 3.5 Sonnet
  • CLAUDE-OPUS - Request was routed to Claude 3 Opus (premium)
  • GPT-4O - Request was routed to OpenAI GPT-4o
  • GEMINI - Request was routed to Google Gemini
  • MISTRAL - Request was routed to Mistral AI

Detailed Analytics

For comprehensive usage analytics, visit your Dashboard to see:

  • Total requests and tokens used
  • Cost breakdown by provider
  • Historical savings trends
  • Usage patterns and peak times

Tips and Best Practices

Maximize Your Savings

  1. Use Auto mode - Let Plexor Labs' routing engine optimize your requests automatically
  2. Start with Eco - For exploration and learning, Eco mode provides maximum savings
  3. Switch modes contextually - Use Quality mode only when you need it for critical work
  4. Be specific - Clear, focused questions often get routed to cheaper models

Get Better Responses

  1. Provide context - Include relevant background information in your first message
  2. Use file attachments - For code questions, attach the actual files rather than pasting
  3. Break down complex tasks - Multiple focused requests often work better than one large request
  4. Force providers when needed - If you know a specific provider works best for your task, select it directly

Keyboard Shortcuts

  • Enter - Send message
  • Shift + Enter - New line in message
Try It Out!
Ready to start? Open the Playground and send your first message. Watch the savings badge as Plexor Labs routes your requests to the optimal provider!

Related Documentation