Use cases
These are the examples we run end to end today. Each one takes an event stream, asks a question about it later, and gets back a ranked answer that explains itself. Every page walks you through the same parts, from how the problem maps onto HeapDB to setup, explanations, benchmarks and the limits we know about.
Building an agent? Start with agent memory: heapdb-memory, our open-source package for recall, with an MCP server and a quickstart.
- AvailableBenchmark: measured
Code changes: likely culprits and review priority
Which merged PR most likely broke main or caused this error spike, and which open PR deserves a careful review?
Read the example: Code changes: likely culprits and review priority → - AvailableNo benchmark yet
Incidents: which change caused the spike?
An alert fired on checkout. Which recent deploy, config change or flag flip most likely caused it?
Read the example: Incidents: which change caused the spike? → - AvailableBenchmark: measured
Product analytics: what goes with outcomes?
Which features, pages and errors go with conversion, churn or support tickets, per user and across users?
Read the example: Product analytics: what goes with outcomes? → - AvailableBenchmark: measured
AI agent memory: recall what matters now
Of everything your agent has seen, which memories are most useful for this turn? We explain every recall and learn from what your agent actually used.
Read the example: AI agent memory: recall what matters now → - AvailableNo benchmark yet
Personal health: why am I bloated?
Bloated at 16:00. Which of today’s meals most likely did it, given what has bothered you before?
Read the example: Personal health: why am I bloated? →
Looking further out? In our RFCs we design HeapDB for other domains, from outbreak tracing to security alert triage, and every one ships with a runnable example.
Have a stream and a question we haven’t covered? Post it on the board. We turn the most-voted ideas into the next examples.
Share your use case