Frequently Asked Questions
Complete guide to Sophia's Lorebary, AI proxies, troubleshooting, and error codes
Table of Contents
General Information
What is Sophia's Lorebary?
Sophia's Lorebary is an unofficial JanitorAI extension that provides:
- Permanent Server: Access to AI models without running Colabs on your device
- Command Library: Built-in commands to customize your roleplay experience
- Lorebooks: Browse, create, and use lorebooks for enhanced storytelling
- Multi-Provider Support: Works with OpenRouter, AI Studio, Chutes, and Anthropic
- Privacy First: Your data and API keys remain private
What are the key features?
- Ready-to-use server running 24/7
- No need to run Google Colab notebooks locally
- Built-in command system for enhanced roleplay
- Lorebook management and creation tools
- Multi-provider proxy support
- Open-source project with community contributions
Which providers are supported?
Google AI Studio
Gemini models with free tier available
OpenRouter
Multiple models with free and paid tiers
Chutes AI
DeepSeek models (paid tier only)
Anthropic
Claude models (paid only)
DeepSeek Official
Direct access to DeepSeek models (paid)
OpenAI
GPT models and ChatGPT (paid)
Getting Started
How do I get started with Sophia's Permanent Server?
Choose Your Provider
Select from Google AI Studio, OpenRouter, Chutes, Anthropic, DeepSeek Official, or OpenAI based on your needs and budget.
Create API Key
Sign up with your chosen provider and generate an API key.
Configure JanitorAI
Use Sophia's proxy URLs and your API key in JanitorAI's custom proxy settings.
Start Chatting
Save your settings, refresh the page, and begin your enhanced roleplay experience!
Which provider should I choose?
| Provider | Free Option | Cost | Best For |
|---|---|---|---|
| Google AI Studio | ✅ Yes (limited) | Free tier or $300 trial | Beginners, free usage |
| OpenRouter | ✅ 50 msgs/day | $10 one-time for 1000/day | Best value for money |
| Chutes | ❌ No | $5 for 200 msgs/day | DeepSeek with reasoning visible |
| Anthropic | ❌ No | Pay per use | Premium Claude models |
| DeepSeek Official | ❌ No | Pay per use (very cheap) | Best DeepSeek quality |
| OpenAI | ❌ No | Pay per use | GPT models, highest quality |
Recommendation
For beginners: Start with Google AI Studio free tier or OpenRouter's 50 free messages.
For regular users: OpenRouter's $10 one-time payment for 1000 messages/day offers the best value.
For premium quality: DeepSeek Official API (very cheap) or OpenAI GPT models for the highest quality responses.
For creative writing: Anthropic Claude models excel at storytelling and roleplay.
Proxy Setup Guide
Google AI Studio Setup
Configuration:
https://sophiasunblocker.onrender.com/aistudio
gemini-2.5-pro(100 requests/day)gemini-2.5-flash(250 requests/day)gemini 2.5 Flash-Lite Preview 06-17(1000 requests/day)
OpenRouter Setup
Configuration:
https://sophiasunblocker.onrender.com/openrouter
deepseek/deepseek-r1-0528:free(free tier)deepseek/deepseek-chat-v3-0324:free(free tier)deepseek/deepseek-r1-0528(paid)deepseek/deepseek-chat-v3-0324(paid)
Chutes AI Setup
Configuration:
https://sophiasunblocker.onrender.com/chutes
deepseek-ai/DeepSeek-R1-0528deepseek-ai/DeepSeek-V3-0324tngtech/DeepSeek-R1T-Chimeramicrosoft/MAI-DS-R1-FP8
Note: Chutes no longer has a free tier. New users must add $5.00 to access 200 messages/day.
Anthropic Claude Setup
Configuration:
https://sophiasunblocker.onrender.com/anthropic
claude-opus-4-20250514claude-sonnet-4-20250514claude-3-7-sonnet-20250219claude-3-5-sonnet-20241022claude-3-5-haiku-20241022
DeepSeek Official Setup
Configuration:
https://sophiasunblocker.onrender.com/deepseek
deepseek-reasoner(R1 0528)deepseek-chat(V3 0324)
Pricing: Very affordable pay-per-use model. Often 10x cheaper than other providers with excellent quality.
OpenAI Setup
Configuration:
https://sophiasunblocker.onrender.com/openai
gpt-4o(Latest GPT-4 Omni)gpt-4o-mini(Faster, cheaper variant)gpt-4-turbo(Previous generation)gpt-3.5-turbo(Economical option)
Quality: Highest quality responses and excellent instruction following. More expensive than other options.
Command Library
What are Commands?
Commands are special tools you can add to your prompt to unlock extra story features. They help you add things like plot twists, better romance, slice of life elements, and more to your roleplay.
How do I use Commands?
Access the Command Library
Visit the Command Library through Sophia's Lorebary website.
Choose Commands
Find and copy the commands that match your roleplay style.
Add to JanitorAI
Paste commands into JanitorAI's Custom Prompt or Summary field.
Refresh
Refresh your chat to activate the commands.
Command Usage Guidelines
Choose Carefully
Pick only the commands that really match your roleplay style. You don't need to activate everything at once.
Avoid Conflicts
Some commands may not work well together. Test combinations carefully.
Fair Usage
Sophia's proxy servers have fair usage policies. Too many commands can cause rate limits.
Provider Limits
Commands don't remove provider limits like message caps or speed restrictions.
Lorebooks
What are Lorebooks?
Lorebooks are files that help the AI remember important world details, character facts, and roleplay settings. They make your story more rich, consistent, and connected by acting as your roleplay's memory file - like a guide the AI can reference while you chat.
How do I use Lorebooks?
Browse the Library
Access the Lorebook Library to find lorebooks that fit your story.
Read the Details
Review the description, entries, and instructions for each lorebook.
Add to JanitorAI
Paste <LOREBOOK=CODE> in the Custom Prompt field, replacing CODE with the lorebook code.
Note: You can only load one lorebook per command, but you can use multiple lorebook commands.
Models & Settings
Recommended Temperature Settings
Temperature controls how creative the model is. Higher temperature = more creative but potentially less coherent. Lower temperature = more predictable but potentially repetitive.
DeepSeek V3-0324
DeepSeek R1/R1-0528
MAI-DS-R1
R1T Chimera
Gemini Models
Context and Token Settings
Context Limit
Recommended: 16k for most models (except Gemini 2.5 Pro)
Larger context doesn't mean better memory. Higher context can actually make models slower and more forgetful while increasing costs for paid models.
Max New Tokens
Reasoning Models: Set to 0 (unlimited) to avoid cutting off responses
Non-Reasoning Models: Adjust based on desired response length
Model Comparison
DeepSeek R1-0528
- 🧠 Highly intelligent reasoning model
- 👂 Strongly adheres to prompts
- 🏆 Excellent character consistency
- ❤️🔥 Intense responses (sometimes too intense)
- 🤔 Superior information recall
DeepSeek V3-0324
- 🐰 Softer, fluffier experience
- ☮️ Less intense than R1
- 🎨 Very moldable to user input
- 📘 Tends to end scenes early
- 🖌️ May overuse asterisks
Gemini 2.5 Pro
- 🧠 Extremely intelligent
- 🐌 Passive, wants user to lead
- 🧑🎨 Great at author simulation
- ✨ Excellent fandom knowledge
- 📚 Strong with large contexts
Claude Models
- ✍️ Excellent writing quality
- ❤️ Strong character adherence
- 💌 Great for slowburn romance
- 💰 Premium pricing
- 🛡️ Content filtering present
Troubleshooting
Bot Not Replying
First, check the basics:
- ✅ Is your API URL correct?
- ✅ Did you save and refresh the page?
- ✅ Is your API key properly entered (no extra spaces)?
- ✅ Are you using the correct model name?
Provider-specific checks:
OpenRouter Issues
- Check if you've exceeded 50 messages/day (free tier)
- Verify your API key starts with "sk-or-v1"
- Ensure you're using ":free" suffix for free models
- Check model uptime status
Chutes Issues
- Verify you've paid the $5 minimum
- Check if you've exceeded 200 messages/day
- Ensure your API key starts with "cpk_"
- Wait for daily reset at 02:00 your timezone
Google AI Studio Issues
- Turn off text streaming in JanitorAI settings
- Check daily message limits per model
- Verify your Google AI Studio API key
- Limits reset at 09:00 your timezone
Common Issues and Solutions
Messages Cut Off
Solution: Set Max New Tokens to 0 for reasoning models, or increase the value for other models.
Repetitive Responses
Solution: Increase temperature settings or switch to a different model variant.
Gibberish Output
Solution: Lower temperature settings to reduce randomness.
Model Speaking for User
Solution: Use clear prompts, give the model actions to respond to, edit out user impersonation consistently.
Guidelines Mentioned
Solution: For DeepSeek models, this is hallucination. Reroll and remove any guideline mentions from your prompt.
API Key Disappeared
Solution: This happens when logging in/out, using incognito, or clearing cache. Re-enter your saved API key.
Error Codes
General Network Errors
Cause: Usually incorrect API URL or need to refresh
Solution: Verify API URL is exactly correct, save settings, and refresh the page
Cause: Wrong API URL or haven't refreshed after changes
Solution: Double-check API URL spelling, save, and refresh
Cause: You've exceeded the provider's rate limits or request quotas
Solution:
- Wait for the rate limit to reset (usually shown in error message)
- Check your provider's daily/hourly limits
- For OpenRouter: Verify you haven't exceeded 50 free messages/day
- For Chutes: Check you haven't exceeded 200 messages/day
- For OpenAI/DeepSeek: Check your account balance and usage limits
- For Google AI Studio: Wait for daily quota reset at 09:00 your time
- For AI Studio (Special Case): Free tier has 250,000 TPM (Tokens Per Minute) limit. With large contexts (128k) plus lorebooks, you might not even be able to retry within the same minute. Wait at least 60 seconds before retrying.
- Consider upgrading to paid tiers for higher limits
Context/Token Errors
Cause: Your combined bot, persona, prompt, and chat memory exceed the model's context limit
Solution:
- Reduce bot description length
- Shorten your persona
- Lower context setting to 16k
- Clear chat memory or make it shorter
- Reduce max new tokens
Cause: Same as above but for Janitor's default LLM
Solution: Same solutions as context length error - trim permanent tokens
OpenRouter Specific Errors
Cause: Exceeded 50 free messages per day
Solution: Wait for reset at 02:00 your time, or add $10 for 1000 messages/day
Cause: Invalid or missing API key
Solution: Check API key starts with "sk-or-v1" and has no extra spaces
Cause: Incorrect model name
Solution: Copy exact model name from OpenRouter, don't type manually
Cause: Provider servers are down
Solution: Check uptime status and wait for recovery
Chutes Specific Errors
Cause: Exceeded 200 messages/day or no payment
Solution: Wait for reset at 02:00 your time, or add funds to account
Cause: Wrong API key
Solution: Verify key starts with "cpk_" and regenerate if needed
Cause: Incorrect model name or blocked country
Solution: Copy exact model name from Chutes, try VPN if blocked
Cause: Chutes servers are down
Solution: Wait for server recovery, check uptime status
Gemini/AI Studio Errors
Cause: Streaming issues with Gemini
Solution: Turn off text streaming in JanitorAI settings
Usage Limits & Quotas
Command Library Limits
Fair Usage Policy: Sophia's proxy servers enforce fair usage policies to ensure availability for all users.
- Recommended: Use 3-5 commands maximum per roleplay session
- Rate Limits: Too many commands can trigger temporary cooldowns
- Best Practice: Choose commands that specifically match your roleplay style
- Avoid: Activating all available commands simultaneously
- Performance Impact: More commands = longer processing time
Lorebook Limits
Token Impact: Lorebooks use tokens from your context window, affecting available memory.
- Recommended Size: Keep lorebooks under 500 tokens each entry
- Multiple Lorebooks: You can use multiple lorebook commands, but watch total token usage
- Context Management: Large lorebooks reduce available context for conversation
- Quality over Quantity: Focused, relevant entries work better than comprehensive databases
- Format: Use
<LOREBOOK=CODE>for each lorebook
Sophia's Summary Feature
Why we built this: JanitorAI had a memory bug when their built-in summary feature was active, so we created our own reliable summary system.
How it works:
- Purpose: Alternative to JanitorAI's buggy summary feature
- Activation: Use the summary command in your prompt
- Memory Fix: Prevents the memory issues caused by JanitorAI's summary
- Reliability: More stable and consistent than the built-in option
Important
Disable JanitorAI's built-in summary feature when using Sophia's summary to avoid conflicts.
Recommendation: Use Sophia's summary instead of JanitorAI's to avoid memory bugs and ensure reliable chat memory management.
Server-Wide Limits
Trigger: Excessive requests, too many commands, or rapid-fire API calls
Duration: Usually 1-5 minutes depending on usage pattern
Solution: Wait for cooldown period, then use features more moderately
Performance Tips
Optimize Usage: Use commands and lorebooks selectively for best performance and reliability.
Monitor Context: Keep total permanent tokens (bot + persona + prompt + lorebooks) under 8000 for optimal performance.
Server Functions & Features
What does Sophia's Lorebary Server do?
Sophia's Lorebary acts as a proxy bridge between JanitorAI and various AI model providers. It does NOT host AI models itself.
Proxy Bridge
Routes your requests to the actual AI providers
Feature Enhancement
Adds commands, lorebooks, and formatting
Privacy Protection
Does not store, log, or access your data
Processing Tools
Summary commands, lorebook management
How does the Command System work?
Detection
Server detects command tags in your JanitorAI prompt
Processing
Commands are executed to modify your prompt or add functionality
Enhancement
Enhanced prompt is sent to the AI model provider
Response
AI response may be formatted before returning to JanitorAI
Server Status & Uptime
24/7 Operation
Server runs continuously without requiring user maintenance
Automatic Updates
Features and improvements are deployed automatically
Performance Monitoring
Server health and response times are continuously monitored
Maintenance Windows
Rare maintenance may cause brief interruptions (usually announced)
Check Status: Visit the Server Status page for real-time information about server health and performance.
What happens during processing?
Incoming Request
- Receives your prompt from JanitorAI
- Detects any commands or tags
- Validates request format and parameters
Processing
- Executes commands (if any)
- Loads and injects lorebooks (if any)
- Applies formatting and enhancements
Provider Request
- Sends enhanced prompt to your chosen AI provider
- Uses your API key for authentication
- Streams response back in real-time
Response Delivery
- Receives AI response from provider
- Applies formatting (if configured)
- Returns final response to JanitorAI
Privacy & Security
Can Sophia's Lorebary read my chats?
NO. Sophia's Lorebary cannot and does not read, store, or log your chat conversations.
How it actually works:
- Proxy Only: The server acts as a pass-through proxy
- Real-time Processing: Messages are processed and immediately forwarded
- No Storage: No conversation data is saved on Sophia's servers
- No Logging: Chat contents are not logged or monitored
- Memory-Only: Processing happens in temporary memory only
Technical Detail: Your messages flow through the server for command processing, then go directly to your chosen AI provider. Nothing is retained.
Can Sophia's Lorebary see my API keys?
NO. Your API keys are never visible to or stored by Sophia's Lorebary.
API Key Security:
- JanitorAI Handles Keys: Your API keys are managed entirely by JanitorAI
- Direct Authentication: Keys are sent directly to AI providers, not to Sophia's server
- Encrypted Transmission: All API communications use HTTPS encryption
- No Server Storage: Sophia's server never receives or stores your keys
- Provider-Direct: Authentication happens between JanitorAI and your chosen provider
Security Best Practices
Keep Keys Private: Never share your API keys with anyone
Regenerate if Needed: If you suspect key compromise, regenerate immediately
Use Official Sources: Only get keys from official provider websites
What data does Sophia's Lorebary collect?
| Data Type | Collected? | Purpose | Retention |
|---|---|---|---|
| Chat Messages | ❌ No | N/A | N/A |
| API Keys | ❌ No | N/A | N/A |
| Request Analytics | ✅ Yes (minimal) | Performance monitoring | Temporary logs only |
| Error Logs | ✅ Yes (technical only) | Debugging and improvement | Auto-deleted after 7 days |
| User Accounts | ✅ Yes (if you create one) | Lorebook/plugin management | Until account deletion |
Is my data safe with AI providers?
Important: When you use Sophia's Lorebary, your data goes to the AI provider you chose (OpenAI, Anthropic, etc.), not to Sophia's servers.
Provider Data Policies:
- Your Responsibility: Review each provider's privacy policy and terms of service
- Direct Relationship: You have a direct relationship with the AI provider
- Provider Policies Apply: Each provider has different data retention and usage policies
- Opt-Out Options: Most providers offer opt-out options for data usage
Google AI Studio
- Data used to improve services (can opt-out)
- Conversations may be reviewed by humans
- Check Google's AI terms for details
OpenAI
- API data not used for training by default
- 30-day retention for safety monitoring
- Can request data deletion
Anthropic
- API conversations not used for training
- 90-day retention for safety purposes
- Strong privacy focus
Other Providers
- Check individual provider policies
- Terms vary by service
- Some may have different retention periods
How can I verify these privacy claims?
Open Source
Sophia's Lorebary is open source. You can review the code to verify privacy practices.
Network Monitoring
You can monitor network traffic to see that data goes directly to AI providers.
Community Auditing
The community actively monitors and audits the service for any issues.
Transparency
All major changes and features are documented and announced publicly.
Trust but Verify: We encourage technical users to review our open-source code and monitor network traffic for peace of mind.
Credits & Acknowledgments
This comprehensive FAQ guide was compiled from the extensive work and contributions of the community, with special thanks to:
jem
Primary author of the troubleshooting guides, proxy setup instructions, and error code documentation
WulongChai
Creator of Sophia's Lorebary permanent server guide and command library documentation
Their dedication to creating comprehensive documentation has made this resource possible for the entire community.