SwiftCache is a high-performance, in-memory data store purpose-built for Agentic AI workflows. It acts as a dedicated “Neural Memory Layer”, enabling agents to efficiently store, retrieve, and manage long-term context and complex reasoning graphs — all with sub-millisecond latency.
SwiftCache aims to become the default memory + orchestration layer for Agentic AI systems, bridging the gap between:
- Stateless LLM execution
- Persistent contextual intelligence
- Real-time observability
A true foundation for building autonomous, scalable, and intelligent agent ecosystems.
SwiftCache follows a dual-interface architecture, cleanly separating high-speed data operations from system observability:
-
⚡ TCP Engine (Port 7777)
A custom binary protocol built on Java NIO for ultra-fast, low-overhead communication between agents. -
📊 REST Dashboard API (Port 8080)
A Spring Boot-based management layer providing real-time metrics, system insights, and memory introspection for frontend dashboards.
The core engine is built using Java 17 and Spring Boot, with a strong focus on thread safety, low latency, and memory efficiency.
SemanticList→ Sliding window buffer for managing LLM context efficientlyReasoningGraph→ Structured representation for tracking multi-step agent logic
- Append-Only File (AOF) ensures data durability across restarts
- Built-in AOF Sweeper compacts logs and removes expired entries automatically
- Tracks CPU usage, heap memory, and thread metrics
- Powered by
com.sun.managementfor deep JVM insights
- TCP session authentication via
AUTHcommand
- JDK 17+
- Maven
Update the authentication password in:
src/main/resources/application.properties
mvn clean install
java -jar target/swiftcache-0.0.1.jar| Command | Parameters | Description |
|---|---|---|
AUTH |
<password> |
Authenticates session (required first command) |
PUSH |
<id> <message> |
Append message to a context |
GRAPH_ADD |
<id> <t_id> <desc> |
Add a task node to reasoning graph |
GRAPH_UPDATE |
<id> <t_id> <STATUS> |
Update task status (COMPLETED / FAILED / PENDING) |
GET_CONTEXT |
<id> |
Retrieve full context messages |
GRAPH_GET |
<id> |
Retrieve reasoning graph |
PING |
— | Health check → returns PONG |
# 1. Authenticate first (Crucial!)
AUTH supersecret
# 2. Check if the engine is breathing
PING
# 3. Create a Chat Agent (SemanticList)
PUSH agent:001 SYSTEM: You are an autonomous researcher.
PUSH agent:001 USER: Analyze the latest market trends.
# 4. Create a Task Workflow (ReasoningGraph)
GRAPH_ADD task:99 setup_env Initialize Docker containers
GRAPH_ADD task:99 fetch_data Scrape competitor pricing
# 5. Update a task status
GRAPH_UPDATE task:99 setup_env COMPLETED
# 6. Retrieve the state
GET_CONTEXT agent:001
GRAPH_GET task:99SwiftCache includes a modern observability dashboard designed as a “War Room for Agents”, emphasizing clarity, density, and real-time feedback.
- Live tracking of Operations Per Second (OPS) and system throughput
- Searchable grid view of all active memory objects in RAM
- Deep inspection panel for analyzing agent states and reasoning flows
- Dark-mode interface with blur effects, neon accents, and micro-interactions
- Optimized for ultra-wide displays and high-density monitoring
- Framework: React 18 + Vite
- Styling: Tailwind CSS
- Animations: Framer Motion
- Charts: Recharts
- Icons: Lucide React
- Node.js 18+
npm install
npm run devEnsure API proxy is configured in vite.config.js:
server: {
proxy: {
'/api/v1': 'http://localhost:8080'
}
}- Binary-safe log of all memory operations
- Deleting this file resets system state
- Human-readable audit log
- Contains timestamps, commands, and authentication attempts
Screenshot of frontend
