Skip to content

AI Assistant (Archie)

Archie is your intelligent assistant, powered by Claude and grounded in your Knowledge Graph. Unlike generic AI chatbots, Archie provides verifiable responses with citations, provenance chains, and contradiction detection.

What Makes Archie Different

Generic AI Chatbots:

  • Generate fluent but unverifiable responses
  • Hallucinate facts confidently
  • No source citations
  • Can't detect contradictions

Archie:

  • Queries Knowledge Graph before responding
  • Every claim includes source citations
  • Detects contradictions symbolically
  • Confidence scores guide trust
  • Full provenance tracking

26 Built-In Tools

Pro+

Archie accesses specialized tools for different tasks:

Document Tools

  • search_documents - Find relevant documents
  • get_document_content - Retrieve full document text
  • analyze_document - Extract entities and insights
  • compare_documents - Side-by-side comparison

Knowledge Graph Tools Enterprise

  • query_knowledge_graph - Search entities and relationships
  • traverse_graph - Follow entity connections
  • validate_claim - Verify facts against the graph
  • detect_contradictions - Find conflicting information

Research Tools Pro+

  • web_search - Search via Tavily or Serper
  • external_validate - Verify via Perplexity Sonar
  • ground_findings - Compare against trusted docs
  • generate_report - Create research reports

Workspace Tools

  • list_workspaces - Show available workspaces
  • create_workspace - Set up new workspace
  • get_workspace_summary - AI-generated overview

Action Tools

  • create_task - Generate to-do items
  • set_reminder - Schedule notifications
  • share_document - Create sharing links

Analytics Tools Team+

  • get_usage_stats - Credit and storage metrics
  • analyze_trends - Document upload patterns
  • compliance_report - Audit trail exports

Grounded Responses

Pro+

Every Archie response is grounded in verifiable sources:

Source of Truth Grounding

When you define Source of Truth collections, Archie:

  1. Queries your trusted document library first
  2. Grounds all responses against approved content
  3. Flags information that conflicts with trusted sources
  4. Provides grounding scores (0-100%)

Example Grounded Response:

You: What's our company policy on remote work?

Archie: According to your Employee Handbook (v3.2, updated
Oct 2025):

Employees may work remotely up to 3 days per week with
manager approval. Full-time remote requires VP approval
and is evaluated case-by-case.

[Source: Employee-Handbook-2025.pdf, Section 4.2, Page 12]
[Grounding Score: 98% - Direct quote from trusted policy]

Knowledge Graph Integration

Enterprise

Archie leverages the full Knowledge Graph for intelligence:

CGR3 Pipeline (Context Graph Reasoning)

  1. Retrieve: Extract relevant entities from query
  2. Rank: Find related claims and evidence
  3. Reason: Generate response with provenance

Contradiction Detection

Archie automatically detects when sources disagree:

You: What's the project deadline?

Archie: ⚠️ CONFLICTING INFORMATION DETECTED

Source 1: March 15, 2026
  - Project-Plan-v1.pdf (uploaded Jan 5, 2026)
  - Confidence: HIGH

Source 2: April 1, 2026
  - Client-Email-Thread.pdf (uploaded Feb 1, 2026)
  - Confidence: HIGH

Recommendation: Verify with project manager - client
appears to have requested extension in recent communication.

Proactive Research

Pro+

Archie can conduct autonomous research:

Web Research

  • Generate targeted search queries
  • Search via Tavily, Serper, or Perplexity
  • Synthesize findings into reports
  • Cross-reference with your documents

Dual Validation

  1. Internal: Ground findings against your Source of Truth
  2. External: Verify via Perplexity Sonar with live data

Research Workflow:

You: "Research AI regulations affecting financial services"

Archie:
  [1/5] Generating search queries...
  [2/5] Searching web sources...
  [3/5] Analyzing 47 results...
  [4/5] Grounding against your compliance docs...
  [5/5] Validating externally...

Report generated: AI-Regulations-Financial-Services.pdf

Key Findings (5):
  ✓ 3 verified against your compliance library
  ⚠️ 1 contradicts internal policy (flagged)
  ℹ️ 1 new regulation not in your docs (validated)

Conversation Features

Multi-Turn Context

  • Maintains conversation history
  • Understands follow-up questions
  • References previous queries
  • Context-aware responses

Streaming Responses

  • Real-time response generation
  • Stage events show progress:
  • knowledge_graph - Querying the graph
  • document_search - Finding relevant docs
  • reasoning - Generating response

Citation Tracking

Every fact includes:

  • Source document name and location
  • Page numbers or sections
  • Confidence score
  • Timestamp of source

MCP Integration

Team+

Extend Archie with external tools via Model Context Protocol:

Supported Integrations

  • GitHub: Create issues, search repos, manage PRs
  • Slack: Send notifications, search messages
  • Databases: Query internal systems
  • Custom APIs: Any MCP-compatible service

Example with MCP:

You: "Create a GitHub issue for the bugs in the QA report"

Archie:
  1. Searches for "QA report" in documents
  2. Extracts bug descriptions using AI
  3. Creates GitHub issues via mcp_github_create_issue
  4. Links issues to source document

Result: Created 3 GitHub issues (#142, #143, #144)

Security & Privacy

Data Isolation

  • Archie only accesses documents you can access
  • Row-Level Security enforced on all queries
  • No cross-tenant information leakage

Audit Logging Team+

  • Complete conversation history
  • Tool invocation tracking
  • Response generation logs
  • Credit usage per query

AI Traffic Logs Team+

  • Timestamp and operation type
  • Prompt preview (500 chars)
  • Response preview (2000 chars)
  • Token usage and estimated cost

Cost Management

Credit Usage:

  • Basic chat: 1-3 credits per message
  • With document search: 2-5 credits
  • With Knowledge Graph queries: 3-7 credits
  • With web research: 5-10 credits

Optimization Tips:

  • Be specific in questions (reduces token usage)
  • Use Source of Truth grounding (prevents re-research)
  • Leverage conversation context (don't repeat info)

Use Cases

Legal Research

"Find all contracts with Acme Corp mentioning indemnification"
"What defenses did we use in similar breach cases?"
"Summarize opposing counsel's recent litigation history"

Compliance

"Are we compliant with the new SEC filing requirements?"
"What documents reference GDPR data processing?"
"Generate audit trail for Q4 document access"

Business Intelligence

"What are our top vendors by invoice volume?"
"Compare Q4 2025 vs Q4 2024 performance metrics"
"Research competitor pricing strategies"

Knowledge Management

"What methodology did we use for the Acme project?"
"Find similar proposals we've won in manufacturing"
"Update me on everything related to Project Phoenix"

Getting Started

Web Interface

  1. Click the Archie chat icon
  2. Ask a question in natural language
  3. Review response with citations
  4. Follow up for clarification

API Access

POST /api/v1/chat/message
{
  "message": "What contracts expire in Q1 2026?",
  "conversation_id": "optional-for-context",
  "options": {
    "use_knowledge_graph": true,
    "include_citations": true
  }
}

View Chat API Docs → See Archie Examples →