FAQ
Common questions about using Polyglot as an AI research playground with persistent memory and knowledge integration.
General Research Questions
What makes Polyglot different from ChatGPT or Claude?
Polyglot is designed as a research environment, not just a chat interface. The key differences:
Persistent Memory: Your conversations and insights persist indefinitely across sessions and model switches
Model Comparison: Switch between GPT, Claude, Gemini, and local models while maintaining context
Research Memory: Key insights are extracted and preserved as searchable memory markers
Knowledge Integration: Upload your documents (RAG) and connect tools (MCP) to enhance AI capabilities
Privacy Control: Everything runs locally with optional sync, so you own your research data
Think of it as the difference between using a notepad (traditional AI chat) versus a research lab with persistent equipment, notes, and accumulated knowledge (Polyglot).
Can I use Polyglot for serious academic research?
Yes, Polyglot is specifically designed for rigorous research workflows:
Reproducible Research: Complete audit trails and exportable research data
Controlled Comparisons: Run identical experiments across different AI models
Citation Tracking: Full provenance of information sources and AI responses
Long-term Projects: Support for research spanning months or years
Collaboration: Team research while preserving individual privacy
Data Integrity: Research memory preserved with cryptographic verification
Many researchers use Polyglot for literature reviews, hypothesis development, model evaluation, and collaborative research projects.
How much does Polyglot cost?
Polyglot itself is free and open-source. You only pay for the AI services you choose to use:
Local Models (Ollama): Completely free, runs on your hardware
OpenAI API: Pay OpenAI's API rates for GPT models
Anthropic API: Pay Anthropic's rates for Claude models
Google AI: Pay Google's rates for Gemini models
Optional Sync Server: Free for personal use, hosting costs for team deployments
The memory management, knowledge integration, and research features are all free.
Memory and Context Questions
How does persistent memory work across AI models?
When you switch models (e.g., from GPT-4 to Claude), Polyglot:
Captures Context: Takes a complete snapshot of conversation history and memory markers
Adapts Format: Translates context to work optimally with the target AI model
Preserves Memory: Ensures all research insights and memory markers transfer intact
Maintains Continuity: You continue the conversation as if no switch occurred
This enables controlled comparative studies where you can run the same prompt across multiple models with identical context.
What are memory markers and why are they important?
Memory markers are extracted insights, decisions, and findings that persist beyond individual conversations:
Research Findings: Key discoveries or conclusions from AI analysis
Methodological Decisions: Choices about research approach or process
Hypotheses: Research hypotheses that evolve over time
Evidence: Supporting or contradicting evidence for research claims
Questions: Important questions to explore in future research
Memory markers transform AI interactions from disposable chats into cumulative research intelligence that builds over time.
How much conversation history can I store?
Storage is limited only by your browser's capacity (typically 1GB+ available):
Individual Conversations: No practical limit on conversation length
Total Conversations: Store thousands of research conversations
Memory Markers: Unlimited memory markers with full search capability
Knowledge Base: Supports gigabytes of research documents
Performance: System remains fast even with large research databases
For team environments, server storage can scale to organizational research needs.
Can I export my research data?
Yes, Polyglot provides comprehensive export capabilities:
Complete Research Export: All conversations, memory markers, and knowledge base
Selective Export: Choose specific projects, conversations, or time ranges
Multiple Formats: JSON, CSV, Markdown, and formatted research reports
Citation-Ready: Exports include complete provenance and citation information
Reproducible: Exported data includes all information needed to reproduce research
Model Integration Questions
Which AI models does Polyglot support?
Cloud Models:
OpenAI: GPT-4o, GPT-4, GPT-3.5-turbo, and future models
Anthropic: Claude Sonnet 4, Claude Opus 4, Claude 3.5 Haiku
Google: Gemini Pro, Gemini Pro Vision
Future Models: Architecture designed to integrate new AI providers easily
Local Models via Ollama:
Meta: Llama 3.2, Llama 2
Mistral: Mistral 7B, Mixtral 8x7B
Code Models: CodeLlama, Code Gemma
Specialized: Research-domain fine-tuned models
Custom Models: Any model supported by Ollama
Can I run Polyglot completely offline?
Yes, with local models via Ollama:
Full Functionality: Complete research environment works offline
Local AI Models: Run Llama, Mistral, and other models on your hardware
Memory Management: All memory features work offline
Knowledge Integration: RAG and local tool integration available offline
No Cloud Dependency: Research continues without internet connectivity
This is particularly valuable for sensitive research or unreliable internet environments.
How do I compare responses across different AI models?
Polyglot makes model comparison straightforward:
Start Conversation: Begin research conversation with one model
Build Context: Develop conversation history and memory markers
Switch Model: Change to different AI model (context transfers automatically)
Run Comparison: Ask same question or provide same prompt
Analyze Results: Compare responses with identical context and memory
The system tracks performance metrics and enables statistical analysis of model differences.
Knowledge Integration Questions
What is RAG and how does it help my research?
RAG (Retrieval-Augmented Generation) integrates your research documents into AI conversations:
Document Upload: Add PDFs, papers, notes, and research materials
Semantic Search: AI automatically finds relevant information from your documents
Grounded Responses: AI answers backed by your specific research materials
Citation Tracking: Know exactly which documents inform each AI response
Knowledge Evolution: Understanding improves as you add more research materials
Instead of generic AI responses, you get answers grounded in your specific research domain and materials.
What types of documents can I upload?
Supported formats include:
PDFs: Research papers, reports, documentation
Text Files: Notes, transcripts, plain text research materials
Markdown: Formatted research notes and documentation
Word Documents: Research drafts and collaborative documents
Future Formats: Architecture supports adding new document types
Documents are processed to preserve research context and enable semantic search across your entire knowledge base.
What is MCP and what tools can I connect?
MCP (Model Context Protocol) connects external tools and data sources to enhance AI capabilities:
Research Tools:
File System Access: AI can read/write research files
Database Connections: Query research databases and datasets
Statistical Tools: R, Python, statistical analysis capabilities
Literature Search: Academic database queries and citation management
Collaboration Tools: Integration with research team systems
Custom Tools:
APIs: Connect to research-specific APIs and data sources
Scripts: Execute custom research scripts and analysis tools
External Services: Integration with institutional research infrastructure
How does knowledge search work?
Polyglot uses advanced semantic search across your knowledge base:
Semantic Understanding: Searches by meaning, not just keywords
Context-Aware: Search results ranked by relevance to current conversation
Cross-Reference: Finds connections between documents and conversations
Research Memory: Integrates with memory markers for comprehensive search
Performance: Sub-second search across gigabytes of research materials
Privacy and Security Questions
Is my research data private?
Yes, Polyglot is designed with privacy as a core principle:
Local-First: All research data stored in your browser by default
No Server Required: Complete functionality without sending data anywhere
Optional Sync: Cloud sync is opt-in with encryption and privacy controls
User-Controlled Keys: You control encryption keys for any synced data
**
Last updated