Complete technical documentation for Vector Knowledge. Learn about architecture, integrations, security, and best practices.
Quick Start
API Reference
Security
Best Practices
Sign up with Google OAuth or email/password. No credit card required for the 24-hour free trial.
Navigate to Settings → Integrations → Connect Gmail. Authorize Vector Knowledge to access your Gmail via OAuth2.
Your emails will automatically begin indexing based on your tier limits. Check processing status in the dashboard.
Ask questions about your emails using natural language. The AI will search your knowledge graph and provide answers with citations.
Frontend: Next.js 15, React 18, TypeScript, Tailwind CSS, Shadcn UI. Backend: Next.js API Routes, Supabase (PostgreSQL + Auth), BullMQ job processing. AI Services: OpenAI (GPT-5, GPT-4o), LangChain orchestration, Pinecone vector database. Integrations: Gmail API (OAuth2), Stripe payments, Redis caching.
1) Gmail OAuth Connection → Secure token management with encrypted storage. 2) Email Ingestion → Tier-based limits with delta synchronization. 3) AI Processing → Knowledge graph extraction + entity relationship mapping using GPT-5. 4) Vector Storage → Pinecone embeddings with metadata filtering. 5) Intelligent Search → Hybrid search (vector + Gmail API fallback). 6) Chat Interface → Streaming responses with citation support.
Enterprise-grade security: OAuth2 tokens encrypted at rest using AES-256, TLS/SSL for all data transmission, CSRF protection and rate limiting, regular security audits and vulnerability scanning, secure cloud infrastructure with 95/100 security score, circuit breaker protection against API failures.
Metadata Only: We store email metadata (sender, recipient, subject, date) and processed embeddings - NOT complete email content. Vector Embeddings: Semantic embeddings stored in Pinecone for similarity search. Knowledge Graphs: Entity relationships stored in Supabase PostgreSQL. Retention: Based on subscription tier (24 hours/6 months/2 years).
Vector Knowledge uses OAuth2 for secure Gmail access. We request the following scopes: gmail.readonly (read email metadata and content), gmail.labels (access folder structure). Tokens are encrypted at rest and refreshed automatically. You can revoke access anytime from Google Account settings or Vector Knowledge settings.
Initial Sync: Fetches emails within your tier limit and retention window. Delta Sync: Real-time sync for new emails (checks every 5 minutes). Folder Selection: By default, syncs Inbox, Sent, and All Mail. Excludes Spam and Trash. Rate Limiting: Complies with Gmail API rate limits (250 quota units/user/second).
Metadata: Sender, recipient, CC, BCC, subject, date, labels, thread ID. Content: Email body (HTML and plain text), attachments (metadata only, not content). Threads: Email conversations are preserved with thread relationships. Labels: Gmail labels are imported and searchable.
Attachments: Attachment metadata is indexed but file contents are not processed. Large Emails: Emails over 10MB are truncated for processing. Calendar Events: Calendar invites are treated as regular emails. Confidential Mode: Emails sent with Gmail confidential mode may have limited access.
Vector Knowledge uses OpenAI embeddings and Pinecone vector search for semantic understanding. Example: Searching 'project deadline' returns results about 'due dates', 'milestones', 'delivery schedules' - even without exact keyword matches. Supports: Natural language queries, temporal queries ('last week', 'Q4 2024'), entity-based queries ('from Sarah', 'about the merger').
Entities Extracted: People (email addresses, names), Companies/Organizations, Topics/Themes, Dates and timeframes, Projects and initiatives. Relationships: Communication patterns (who emails whom), Topic associations (which topics are discussed together), Temporal patterns (when communications occur). Visualization: Interactive graph showing your professional network and communication dynamics.
Model: GPT-5 (production), GPT-3.5-turbo (development). Features: Streaming responses for real-time interaction, citation support (links to source emails), context-aware conversations (maintains chat history), semantic retrieval (finds relevant emails first, then generates answers). Prompts: Pro/Premium tiers have unlimited prompts. Free trial limited to 25 prompts total.
AI automatically extracts structured entities from emails: Person entities (names, email addresses, job titles), Company entities (organization names, domains), Topic entities (key themes and subjects), Date entities (deadlines, meetings, events). These entities power search filters and knowledge graph visualization.
At Rest: OAuth2 tokens encrypted with AES-256, database encrypted with Supabase's transparent data encryption, vector embeddings encrypted in Pinecone. In Transit: TLS 1.3 for all API communication, HTTPS enforced for all web traffic, secure WebSocket connections for real-time features.
OAuth2 Flow: Secure delegation to Gmail without storing your Google password. Session Management: JWT tokens with configurable expiration, secure session cookies (HttpOnly, Secure, SameSite). Multi-Factor Authentication: Supported through Supabase Auth. Role-Based Access: Granular permissions for different user roles.
GDPR Compliant: Full data subject rights (access, rectification, deletion, portability), data processing agreements with all subprocessors, EU-US data transfer mechanisms (Standard Contractual Clauses). SOC 2 Type II: Infrastructure partners (Supabase, Pinecone, Vercel) maintain SOC 2 certification. Regular Audits: Quarterly security audits and penetration testing.
Retention Periods: Free (24 hours), Pro (6 months), Premium (2 years). Automatic Purging: Old data automatically deleted per tier limits. Account Deletion: Within 30 days of account deletion request, all data marked for deletion. Backups purged within 90 days. OAuth tokens immediately revoked.
We're developing a REST API for programmatic access to Vector Knowledge features. Planned endpoints include: Email search and retrieval, AI chat integration, Knowledge graph queries, Usage statistics and analytics. API keys will be available in Pro and Premium tiers. Documentation will be available at /api-docs when launched.
Receive real-time notifications for: New emails indexed, AI processing completed, Usage limit warnings, Subscription changes. Webhook endpoints can be configured in Settings → Developer.
Official SDKs will be available for: JavaScript/TypeScript (Node.js and browser), Python, Go. Community SDKs are welcome - contact tim@vectorknowledge.com for support.
Web Application: 100 requests per minute per user. API (when available): 1,000 requests per hour (Pro), 5,000 requests per hour (Premium). AI Chat: Subject to OpenAI rate limits (generally very high for our use case). Email Sync: Compliant with Gmail API limits (no user action required).
Use natural language: 'emails about the Q4 budget meeting' works better than keyword stuffing. Be specific about timeframes: 'last week', 'Q4 2024', 'in January'. Specify senders/recipients: 'from Sarah about project alpha'. Use topic-based queries: 'discussions about pricing' finds all pricing-related emails.
Regular Cleanup: Archive or delete old emails in Gmail to stay within tier limits. Prioritize Important Emails: Vector Knowledge indexes most recent emails first within tier limits. Upgrade When Needed: Monitor usage in dashboard; upgrade before hitting limits. Folder Management: Exclude low-priority folders to maximize valuable email indexing.
Sensitive Emails: Vector Knowledge processes all emails - be mindful of highly sensitive content. Access Control: Use strong passwords and enable 2FA. Don't share accounts. Review Permissions: Periodically review connected apps in Google Account settings. Disconnect When Not Using: You can disconnect Gmail anytime and reconnect when needed.
Check Help Center: Most common questions answered at /help. Contact Support: tim@vectorknowledge.com for general inquiries (24-48 hour response). Priority Support: Pro and Premium users get priority email support. Emergency Security: tim@vectorknowledge.com for urgent security issues (24/7).
Supported Browsers: Chrome 100+, Firefox 100+, Safari 15+, Edge 100+. Not Supported: Internet Explorer. Recommendations: Use Chrome or Firefox for best performance. Enable JavaScript and cookies. Disable ad blockers that may interfere with API calls.
Bandwidth: Minimum 5 Mbps for email sync and AI chat. Recommended 10+ Mbps for optimal experience. Ports: HTTPS (443) must be accessible. WebSocket support required for real-time features. Firewall: Allow connections to: *.vectorknowledge.com, *.supabase.co, *.pinecone.io.
Account Type: Personal Gmail or Google Workspace accounts supported. Permissions: Must have permission to authorize third-party apps (check with IT admin for Workspace). 2-Step Verification: Supported (may require app password in some cases). IMAP Access: Must be enabled in Gmail settings.
Clear browser cache regularly to prevent slowdowns. Close unused tabs to free up memory. Use incognito mode if experiencing issues (to isolate extension conflicts). Update browser to latest version. Disable unnecessary browser extensions that may interfere.
We're developing a comprehensive REST API for programmatic access to Vector Knowledge. API keys will be available for Pro and Premium tiers.
Interested in API access? Contact us