An AI Operating System for modern software engineering teams.
Pulse is an AI-native Project Intelligence Platform designed to move beyond traditional project management. Instead of simply tracking tasks, Pulse understands project context, reasons over historical decisions, automates engineering workflows, and assists teams through specialized AI agents.
Unlike conventional tools that treat AI as an add-on chatbot, Pulse integrates AI into the core architecture through an event-driven system, semantic memory, hybrid retrieval, human approval workflows, and multi-agent orchestration.
Modern engineering teams generate enormous amounts of information:
- Tasks
- Pull Requests
- Architecture Decisions
- Sprint plans
- Meeting notes
- Comments
- Documentation
- Commits
- Deployment logs
Most of this knowledge becomes fragmented across multiple tools.
Pulse transforms these disconnected events into an intelligent knowledge system capable of answering questions, explaining historical decisions, predicting project risks, and automating repetitive engineering work.
Pulse introduces an AI Operating System built around specialized agents rather than a single general-purpose assistant.
Available personas include:
- Planner
- Architect
- Developer
- Reviewer
- QA
- Manager
- Researcher
Each agent maintains:
- Dedicated system instructions
- Model configuration
- Tool access
- Context retrieval
- Evaluation metrics
- Confidence scoring
Requests are routed through an Agent Orchestrator which selects the appropriate persona, gathers project context, invokes tools when necessary, and returns the final response.
Pulse enables AI agents to perform structured actions instead of producing text alone.
Examples include:
- Create Sprint
- Create Task
- Generate Architecture Decision Record
- Estimate Sprint Scope
- Prioritize Backlog
High-impact actions are protected through a Human Approval Queue before execution.
AI-generated changes are not executed automatically.
Instead, Pulse intercepts sensitive actions and creates pending proposals.
Reviewers can:
- Inspect generated parameters
- Approve execution
- Reject requests
- Audit previous proposals
This keeps humans in control while still benefiting from AI automation.
Engineering Memory is Pulse's signature capability.
Instead of relying only on chat history, Pulse retrieves project knowledge from multiple sources.
When a user asks:
"Why did we migrate from MongoDB to PostgreSQL?"
Pulse searches:
- Architecture Decision Records
- Timeline Events
- Audit Logs
- Tasks
- Related project history
The retrieved context is synthesized into a coherent explanation describing:
- Decision rationale
- Authors
- Timeline
- Related implementation work
- Outcome
This allows the platform to explain engineering decisions rather than simply recalling conversations.
Pulse combines multiple retrieval strategies.
User Query
↓
Keyword Search
Vector Similarity Search
↓
Result Fusion
↓
LLM Re-ranking
↓
Context Builder
↓
Final AI Response
This hybrid pipeline improves retrieval quality over vector-only approaches.
Pulse maintains relationships between engineering entities.
Nodes include:
- Organizations
- Workspaces
- Projects
- Sprints
- Tasks
- Developers
- Pull Requests
- Documents
- ADRs
- AI Agents
Relationships are visualized through an interactive graph, enabling exploration of project dependencies and engineering history.
Everything inside Pulse is modeled as an event.
Examples:
- Task Created
- Sprint Updated
- Pull Request Merged
- Proposal Approved
- AI Tool Executed
- Memory Updated
Events are published to a central Event Bus where subscribers update independent systems including:
- Timeline
- Analytics
- Notifications
- Engineering Memory
- Project Wrapped
This decoupled architecture allows new capabilities to subscribe to events without modifying existing modules.
Inspired by annual productivity summaries, Project Wrapped transforms engineering activity into an interactive narrative.
Metrics include:
- Tasks completed
- Longest productivity streak
- Peak working hours
- AI usage
- Sprint achievements
- Code review statistics
- Project milestones
Wrapped is generated from historical events rather than manually tracked counters.
Pulse tracks operational metrics across the platform.
Examples include:
- AI latency
- Token usage
- Estimated inference cost
- Cache statistics
- Tool execution metrics
- Event throughput
These insights provide visibility into both system performance and AI behavior.
Architecture Decision Records document important technical decisions.
The Architect Agent can draft proposals which are reviewed, approved, or archived.
This creates a searchable history explaining why significant architectural changes occurred.
- Next.js 15 (App Router)
- TypeScript
- Tailwind CSS v4
- React
- Motion
- Recharts
- Next.js Route Handlers
- Supabase PostgreSQL
- pgvector
- Row-Level Security (RLS)
- OpenRouter
- Multi-Agent Orchestration
- Tool Registry
- Hybrid RAG
- Engineering Memory
- Vercel
- Supabase
- Event Bus
- Audit Trail
- Vitest
Experience Layer
↓
Application Layer
↓
AI Operating System
↓
Knowledge Layer
↓
Automation Layer
↓
Infrastructure Layer
- Dashboard
- Projects
- Tasks
- AI Copilot
- Wrapped
- Timeline
- Analytics
- Knowledge Graph
- Guardrails
- Engineering Memory
- Planner
- Architect
- Developer
- Reviewer
- QA
- Manager
- Researcher
- Hybrid Retrieval
- Semantic Search
- Engineering Memory
- Knowledge Graph
- Audit Trail
- Event Bus
- Workflow Integrations
- Notifications
- External Webhooks
- Next.js
- Supabase
- PostgreSQL
- pgvector
- Vercel
src/
├── app/
├── components/
├── agents/
├── lib/
├── memory/
├── rag/
├── toolRegistry/
├── analytics/
├── observability/
├── timeline/
├── wrapped/
├── knowledge/
├── guardrails/
├── services/
├── tests/
└── docs/
- Clone the repository.
- Install dependencies.
- Configure environment variables.
- Run database migrations.
- Start the development server.
- Visit the local application.
Production builds can be generated using the standard Next.js build process.
Pulse is a portfolio and research platform demonstrating modern AI-assisted software engineering concepts.
Implemented capabilities include:
- Multi-agent orchestration
- AI tool execution
- Human approval workflows
- Engineering Memory
- Hybrid retrieval
- Knowledge Graph visualization
- Event-driven architecture
- Project Wrapped
- Observability dashboards
- Architecture Decision Records
- Comprehensive documentation
- Automated testing
Some predictive analytics components are currently implemented as prototype pipelines for architectural demonstration and are documented accordingly.
Potential future work includes:
- Training project-specific machine learning models for delivery forecasting
- Expanded workflow automation integrations
- Enhanced collaborative editing
- Additional engineering analytics
- Richer AI evaluation benchmarks
- Production-scale distributed event processing
This repository is intended as an educational and portfolio project demonstrating modern software architecture, AI integration, and engineering best practices.