Build an agent
Start with the Agent Memory overview, then use the Python or HTTP reference for implementation details.
Documentation
Use this page as the shortest path into the docs. Pick the workflow that matches what you are trying to do.
Start here
Section titled “Start here” Quick Start Start with SQL, or jump into the Agent Memory Python loop.
Agent Memory Build an agent flow with time-series evidence, context retrieval, and prompt cache.
Physical AI & ROS 2 Ingest robot telemetry, typed profiles, replay bags, and query sensor timelines.
Logistics & Edge Automation Query AGV, RFID, sorter, cold-chain, and geofence workflows.
HTTP API Use SQL endpoints, Agent Memory endpoints, auth, metrics, and admin APIs.
Python API Query data, use dataframe helpers, and call memory/cache helpers from Python.
Choose a path
Section titled “Choose a path”Run time-series workloads
Learn the SQL surface, ingestion options, benchmarks, and feed handlers.
Connect ROS 2
Build the optional ROS 2 bridge, smoke-test Jazzy packages, subscribe to scalar or typed-profile topics, and import or replay rosbag2 data.
Deploy and operate
Move from a local server to Docker, production deployments, Kubernetes, and security operations.
Evaluate fit
Compare ZeptoDB against kdb+, ClickHouse, InfluxDB, and TimescaleDB before choosing an architecture.
Common tasks
Section titled “Common tasks”| Task | Go to |
|---|---|
| Start a local server | Quick Start |
| Configure Agent Memory persistence | HTTP API Reference |
| Write an agent context flow | Agent Memory |
| Query from Python | Python API Reference |
| Use SQL temporal functions | SQL Reference |
| Use spatial distance and geofence predicates | SQL Reference: Spatial Functions |
| Ingest Arrow IPC batches over HTTP | HTTP Reference: Arrow IPC Ingest |
| Ingest MessagePack column batches over HTTP | HTTP Reference: MessagePack Ingest |
| Connect Telegraf input plugins to ZeptoDB | Telegraf Output |
| Consume AWS Kinesis streams | C++ Reference: Feed Handlers |
| Enable ROS 2 topic ingest | ROS 2 Setup and Smoke Test |
| Deploy ZeptoDB at a ROS 2 edge node | ROS 2 Edge Deployment |
Use Ros2Consumer from C++ | C++ Reference: ROS 2 connector |
Use OpcUaConsumer or OpcUaServer from C++ | C++ Reference: OPC-UA Consumer |
| Inspect cluster-wide Agent Memory stats | HTTP API Reference |
| Deploy with Docker | Docker Deployment |
| Run in production | Production Deployment |
| Follow release branch policy | Branch Release Policy |
| Cut and validate a release | Release Process |
| Set up SSO/RBAC/audit | Security Operations Guide |
| Review latency numbers | Benchmarks |
Current Agent Memory operating model
Section titled “Current Agent Memory operating model”Agent Memory supports routed writes, point reads, fan-out memory search/context assembly, semantic-cache fan-out, replica WAL durability policy, delete and eviction tombstones, tenant quotas, local stats, cluster-scoped stats, and owner-failover reporting.
Shard migration dual-write/catch-up remains future work. For implementation details, use HTTP API Reference, Python API Reference, and Agent Memory for AI Agents.