Real-World Use Cases

These tutorials demonstrate how to combine OpenContext features to build practical, production-ready applications. Each use case shows end-to-end scenarios that integrate multiple concepts.

Available Use Cases

1. Personal Memory Assistant

Build an AI assistant that remembers user preferences, stores notes, and shows how thinking evolves over time.

Features demonstrated:

• Remembering preferences with metadata
• Storing and searching notes
• Time-travel through personal thoughts
• Updating and improving memories

Best for: Personal productivity tools, note-taking apps, memory-aware assistants

Code example: examples/src/tutorials/use-cases/30-personal-memory-assistant.ts

---

2. Customer Support Agent

Create a support bot that remembers customer history, detects repeat issues, and provides personalized service.

Features demonstrated:

• Multi-user memory management
• Customer profile and interaction tracking
• Cross-platform memory unification
• Temporal queries for repeat issue detection
• Batch data import

Best for: Customer service platforms, helpdesk systems, CRM integrations

Code example: examples/src/tutorials/use-cases/31-customer-support-agent.ts

---

3. Research Knowledge Tracker

Track academic findings, citations, and how understanding evolves as new research emerges.

Features demonstrated:

• Storing findings with rich metadata
• Linking related research via themes
• Time-travel through knowledge states
• Synthesizing multiple findings
• Citation-aware search

Best for: Research tools, academic platforms, knowledge management systems

Code example: examples/src/tutorials/use-cases/32-research-knowledge-tracker.ts

---

4. Customer Health Scoring

Identify churn-risk accounts in real time by wiring the distill + derive primitives into a customer success pipeline. Each interaction flows through per-message entity extraction, per-window fact derivation, and per-search signals — surfacing a numeric health score plus human-readable signals that a Loop-engine schedule forwards to the CS team.

Features demonstrated:

• distill for per-message product / pain-point entity extraction
• derive over a 21-day window — all four DerivedKind shapes (summary, frequency, contradiction_candidate, temporal_trend)
• Per-channel signals.entity on search results so CSMs see why a customer match surfaced
• Best-effort, opt-in LLM contract — the demo uses rule-based stubs that swap out for hosted LLMs in production

Best for: Customer success platforms, churn-prevention tooling, CRM enrichment

Code example: examples/src/tutorials/use-cases/35-customer-health-scoring.ts

Running Use Case Examples

Each use case includes a runnable TypeScript example. To run:

cd /path/to/opencontext/examples
pnpm install
node --experimental-strip-types src/tutorials/use-cases/XX-*.ts

Replace XX-*.ts with the specific example file name.

Prerequisites

Before working with use cases, complete:

1. Getting Started - Installation and first API call
2. User Guide - Core concepts and the four verbs
3. Developer Guide - Integration patterns

Learning Path

┌─────────────────────────────────────────────────────────────┐
│                    Your Journey                              │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  1. 📖 Core Tutorials                                      │
│     → Getting Started, User Guide, Developer Guide          │
│                                                             │
│  2. 🚀 Advanced Topics                                     │
│     → Advanced Usage, Best Practices                       │
│                                                             │
│  3. 🏗️ Real-World Use Cases                               │
│     → Personal Memory Assistant (individual scale)          │
│     → Customer Support Agent (multi-user scale)             │
│     → Research Tracker (knowledge evolution)                │
│     → Customer Health Scoring (real-time CS pipeline)       │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Key Patterns Across Use Cases

Metadata Strategies

All use cases leverage metadata for advanced filtering:

metadata: {
  type: "note|ticket|finding",
  category: "work|login|llm-architecture",
  tags: ["tag1", "tag2"],
  // ... custom fields
}

Temporal Queries

Time-travel is used differently per scenario:

• Personal Memory: Track how your thinking changes
• Support Agent: Detect repeat issues over time
• Research: Show knowledge state at specific points
• Health Scoring: Roll up 21-day windows into summary / frequency / contradiction / trend signals

Batch Operations

For data import and bulk operations:

await messages.storeMessages(largeBatch);

Choosing Your Use Case

If you're building...	Start with
Personal productivity app	Personal Memory Assistant
Customer service tool	Customer Support Agent
Research/academic tool	Research Knowledge Tracker
Customer success / churn-prevention tool	Customer Health Scoring

Contributing

Have a use case to share? Contributions welcome!

1. Fork the repository
2. Add your use case to docs/tutorials/use-cases/
3. Create the matching example in examples/src/tutorials/use-cases/
4. Update this README

See CONTRIBUTING.md for guidelines.

---

Next: Choose a use case above to explore!