Agent Memory

Memory is the foundation of agents; governance is the prerequisite of memory. Lingdian turns conversations, diagnoses, and object relations into confirmable assets — resume across sessions, recall by permission, get sharper on your site over time.

Asset types map to Lingdian session engine, expert memory, and device/station brains: refine by layer → govern → assemble per task.

Back to Lingdian AI.

Asset types

Session

Session memory

Long chats compress with provenance; drill back to raw messages when needed.

Expert

Expert memory

Candidates from chat/diagnosis; confirm → own by expert → recall across sessions.

Object Brain

Device / station brain

Relations and diagnosis history hang on the object for the next shift.

Knowledge

Knowledge & Skills

Docs vs case conclusions; assemble per project with inheritance and isolation.

Layers

L3

Stable persona / confirmed factsLong-lived preferences, constraints, confirmed conclusions.

L2

Scenario & decision threadThis commissioning / this alarm handling context.

L1

Atomic facts & eventsRegister findings, phenomena, evidence, diagnosis rows.

L0

Raw chat & dual-track log refsFull messages and provenance after compression.

Governance

Ownership · confirm · reader scope · TTL / merge / revoke · assemble only permitted assets for the current task.

Next

Overview · AI Role · AI Collaboration