CH 09 · Data Modeling for Agents · 01/12
Read as one pageData Modeling for Agents
Model conversations as event logs, version every artifact the agent produces, and track long-running work with explicit status fields.
12 slides · 5 min read
In this chapter
- 01Conversations as event logs, not transcripts
- 02Persist at the edges of the turn
- 03Artifacts as versioned, immutable records
- 04Modeling fine-grained changes
- 05State machines with explicit status fields
- 06The snapshot-vs-current reconciliation
- 07Sharing and multi-tenancy in the schema
- 08Metering and limits
- 09Caching external calls
- 10Choosing a store
- 11What a good agent schema reveals
An agent's data model is easy to underestimate and expensive to get wrong. It must capture conversations richly enough to replay them, version the artifacts the agent produces, preserve an audit trail, and track the state of long-running work. This chapter covers the modeling patterns that recur across agents, independent of any particular database.
Finished this idea?
Your learning progress stays in this browser.