Tembory Agent Memory icon

Tembory Agent Memory

Recommended Tembory memory for n8n AI Agents with stable node identity for the v2 experience

Tembory Agent Memory

AI-generated

Summary

Prepare and supply conversation memory context to an n8n AI Agent, then capture the agent’s messages and tool activity for later turns. The selected preset controls context composition; optional pgvector recall adds archived conversation evidence when an Embeddings connection is available.

Inputs

  • pgvectorEnabled — Enable historical recall from Tembory PostgreSQL; requires an Embeddings connection. When enabled, also configure embeddingSpace and optionally vectorRetentionDays (1–365 days; default 90).
  • threadId — Stable conversation identifier used to load and save thread memory. Defaults to expressions that look for thread, conversation, contact, customer, sender, or session IDs, then the execution ID.
  • project — Workflow namespace for separating clients or workspaces; defaults to available project or account fields.
  • query — Current user message used to build operational context and retrieve relevant memories. Defaults to common chat and webhook message fields.
  • operationPreset — Context mode: recommended Memória Inteligente, Rápido e Econômico, Mini/SLM Orientado, or Auditoria Completa. Defaults to the recommended mode.
  • persistencePolicy — In version 2, choose whether to stop if durable persistence is not confirmed (required, default) or continue while reporting the risk (best effort).
  • contextMaxChars — Approximate maximum context size; defaults to 20,000 characters.
  • recentMessagesLastN — Number of recent messages considered; defaults to 50.
  • debugVisualOutput — Show evidence, tool history, and context health in execution output. Defaults to enabled in version 1 and disabled in version 2.
  • showAdvanced — Reveal technical controls for retrieval mode, memory key, payload format, and detailed context, summary, tool-history, and audit settings.

Output shape

An n8n AI memory connection, with optional execution output containing context diagnostics and evidence.

The memory connection supplies prepared context to the AI Agent and records conversation messages and captured tool activity for subsequent turns. Advanced settings can change which context sections are included and how they are formatted. With pgvector enabled, archived evidence may be added to the context; it is presented as untrusted conversation data.

Links

Discussion