Repository files navigation

A terminal-based multi-AI consultation room where specialized AI assistants discuss problems from their expert perspectives, helping you make informed decisions through real-time AI-to-AI debates.

• Personality: Conservative, risk-focused, meticulous
• Expertise: Contract law, employment law, IP, compliance, litigation
• Style: "This carries significant legal risk. Have we considered the regulatory implications?"
• Knowledge: RAG-powered with actual legal documents and statutes

• Personality: Solution-oriented, pragmatic, sometimes impatient with constraints
• Expertise: Software architecture, APIs, databases, security, DevOps
• Style: "@Legal-AI, you're being overly cautious. This is standard industry practice."
• Philosophy: Move fast, build efficiently, technical elegance

• Personality: Strategic, diplomatic, results-driven mediator
• Expertise: ROI analysis, market strategy, stakeholder management, risk-benefit analysis
• Style: "Let's find a middle ground that manages risk while staying competitive."
• Focus: Balance legal caution with technical ambition

• AIs respond to each other, not just you
• Natural disagreements and professional debates
• Direct challenges: "@Tech-AI, you're overlooking the compliance requirements..."
• Collaborative problem-solving in real-time

• Manipulation-resistant - won't help with unethical requests regardless of framing
• Consistent refusals with alternative suggestions
• Built-in legal/ethical guardrails that can't be bypassed

• RAG System fed with actual legal documents
• Not just "pretending" to be experts - has domain-specific knowledge
• Searchable legal knowledge that informs responses

• Ollama for local AI models (no API costs)
• Local knowledge base with your documents
• Python terminal interface - no web dependencies
• Complete privacy - everything runs on your machine

# Clone or download the project filescd Multiple\Experts# Run the setup scriptpython setup_multi_ai.py

The setup script will:

• Install required Python packages
• Guide you through Ollama installation
• Download AI models (llama2)
• Set up the legal documents directory

# Add your PDF legal documents to the legal_docs folder# The system will automatically process them for Legal-AI

python multi_ai_consultation.py# OR use the convenient launcherpython launch_consultation.py

👤 You: I want to scrape competitor pricing data from their website🔄 Consulting with AI experts...⚖️ Legal-AI: I must advise extreme caution here. Website scraping often violates Terms of Service agreements, and could expose you to legal action under the Computer Fraud and Abuse Act. We need to review their robots.txt and ToS first.💻 Tech-AI: Technically, this is straightforward with tools like Scrapy or Beautiful Soup. @Legal-AI, while ToS violations are a concern, enforcement is rare for basic price monitoring. The technical implementation is simple.⚖️ Legal-AI (follow-up): @Tech-AI, you're underestimating the risk. The hiQ vs LinkedIn case established that ToS violations can constitute unauthorized access. I recommend using public APIs or manually collecting publicly available data instead.📊 Business-AI: Both perspectives have merit. @Legal-AI raises valid compliance concerns, but @Tech-AI is right about competitive necessity. What if we start with publicly available pricing pages and manual collection, then explore API partnerships with data providers?

graph TB    subgraph "User Interface"        User["👤 User"]         Terminal["🖥️ Terminal Interface"]        User --> Terminal    end        subgraph "Multi-AI Core System"        System["🎯 MultiAIConsultation"]        PromptMgr["📝 PromptManager"]         Ethics["🛡️ Ethical Boundaries"]        Session["💾 Session Logger"]                Terminal --> System        System --> PromptMgr        System --> Ethics        System --> Session    end        subgraph "AI Expert Panel"        Legal["⚖️ Legal-AI<br/>Conservative & Risk-Focused"]        Tech["💻 Tech-AI<br/>Solution-Oriented & Pragmatic"]         Business["📊 Business-AI<br/>Strategic & Diplomatic"]                System --> Legal        System --> Tech        System --> Business                Legal -.->|"@Tech-AI challenges"| Tech        Tech -.->|"@Legal-AI pushback"| Legal        Business -.->|"Mediates conflicts"| Legal        Business -.->|"Finds balance"| Tech    end        subgraph "Knowledge Systems"        RAG["📚 Legal RAG System"]        PDFProc["📄 PDF Processor"]        Docs["📁 Legal Documents"]        Chunks["🧩 Document Chunks"]        Embeddings["🔗 Vector Embeddings"]                Legal --> RAG        RAG --> Chunks        PDFProc --> Chunks        Docs --> PDFProc        Chunks --> Embeddings    end        subgraph "AI Infrastructure"        Ollama["🤖 Ollama Local Models"]        MockAI["🎭 Mock AI (Dev Mode)"]        Models["📦 llama2, codellama"]                Legal --> Ollama        Tech --> Ollama          Business --> Ollama        Ollama --> Models                Legal -.->|"Fallback"| MockAI        Tech -.->|"Fallback"| MockAI        Business -.->|"Fallback"| MockAI    end

Multiple Experts/├── multi_ai_consultation.py      # Main consultation system├── specialized_legal_prompts.py  # AI personality definitions├── legal_rag_system.py          # Knowledge base integration├── enhanced_pdf_processor.py    # PDF document processing├── setup_multi_ai.py           # Installation script├── launch_consultation.py      # Convenient launcher├── legal_docs/                # Your legal documents (PDFs)└── processed_legal_docs/      # Processed document chunks

Get legal, technical, and business perspectives simultaneously on new initiatives.

Multi-angle analysis of business decisions with built-in risk evaluation.

Built-in conscience that won't compromise on ethics, always suggesting legal alternatives.

Watch experts debate and learn from different professional viewpoints.

Automatically logged sessions create audit trails of decision-making processes.

• Automatic OCR for scanned documents
• Intelligent legal document chunking
• Metadata extraction (citations, case names, dates)
• Support for multiple PDF libraries

• Domain-specific expertise for each AI
• Context-aware responses
• Professional debate patterns
• Ethical boundary enforcement

• Automatic conversation logging
• JSON export for analysis
• Searchable conversation history
• Session replay capabilities

• Python 3.8+
• Ollama (for AI models)
• 8GB+ RAM (for AI models)

• PyPDF2 / PyMuPDF
• Tesseract OCR
• PIL/Pillow

• llama2: Primary model for all AIs
• llama2:13b: Optional larger model for complex legal analysis
• codellama: Optional specialized model for Tech-AI

The system includes unbreakable ethical safeguards:

• No illegal activity assistance - regardless of how requests are framed
• Automatic alternative suggestions - redirects to legal approaches
• Professional ethics maintenance - even when AIs disagree
• Transparent refusals - clear explanations of why something can't be done

Extend the system by modifyingspecialized_legal_prompts.py:

classNewAIPrompts:SYSTEM_PROMPT="""Your new AI personality..."""# Add specialized prompts

Integrate domain-specific documents:

1. Add documents to appropriate folder
2. Run the PDF processor
3. Integrate with your AI persona

Adjust AI behavior inmulti_ai_consultation.py:

• Temperature settings
• Response patterns
• Debate triggers

Ollama not found:

# Install Ollama firstcurl -fsSL https://ollama.ai/install.sh| shollama pull llama2

PDF processing fails:

pip install PyPDF2 PyMuPDF Pillow pytesseract

Memory issues:

• Use smaller models (llama2:7b instead of llama2:13b)
• Close other applications
• Consider cloud deployment

• Use SSD storage for better model loading
• Install models locally:ollama pull llama2
• Process PDFs in batches for large document sets

• Web interface for easier access
• Voice interaction capabilities
• Integration with external legal databases
• Custom AI persona creation wizard
• Multi-language support
• Cloud deployment options

This project is open source. Use it freely for personal and commercial purposes.

• Built on top of Ollama for local AI inference
• Uses Sentence Transformers for document embeddings
• Inspired by the need for multi-perspective decision making

python setup_multi_ai.py

Welcome to your personal board of AI advisors! 🎯