Skip to main content

Overview

Episodes are chronological records of events — conversations, operations, state changes, and observations. They form the foundational layer of Neuradex’s memory system. For why structured memory matters over simple chunk search, see the concept overview.

Fields

FieldTypeDescription
idstringUnique identifier
contentstringEpisode content
episodeTypestringType of episode
actorTypestringWho performed the action
actorIdstring?Actor identifier
actorNamestring?Actor display name
channelstringSource channel
sessionIdstring?Session grouping ID
parentEpisodeIdstring?Parent episode (for Q&A threading)
scopeTypestringScope type (project)
scopeIdstringScope identifier
occurredAtstringWhen the event occurred
createdAtstringCreation timestamp

Episode Types

Each episode is classified by its type, describing the nature of the recorded event.
TypeDescription
questionUser question
answerAgent or system answer
actionOperation performed
observationState change observed

Actor Types

The actor type indicates who or what generated the episode.
TypeDescription
userHuman user
agentAI agent or assistant
systemSystem-generated

Channels

Episodes can originate from various channels, each representing a different integration point.
ChannelDescription
librarianLibrarian conversations
chatbotChatbot widget conversations
apiDirect SDK/API calls
mcpAI agent operations via MCP
slackSlack integration messages

Session Management

Episodes within the same session are grouped by sessionId. This allows you to retrieve all events that occurred within a single interaction or workflow. Use parentEpisodeId to thread questions and answers together. For example, an answer episode can reference the question episode it responds to, forming a clear Q&A pair.

Data Types

Concept overview of all data types

Episodes API

Manage episodes with the SDK