Ejentum icon

Ejentum

Retrieve a task-matched cognitive operation from the Ejentum Reasoning Harness and inject it into the next LLM step. Each operation is engineered in two layers: a natural-language procedure plus an executable reasoning topology (graph DAG with gates, parallel branches, and meta-cognitive exits). Eight modes: four dynamic (reasoning, code, anti-deception, memory) plus four adaptive (adaptive-reasoning, adaptive-code, adaptive-anti-deception, adaptive-memory) that pre-fit the operation to your task via an adapter LLM. Dynamic available on all tiers including the 30-day free trial; adaptive requires Go or Super.

Actions8

Memory Harness

AI-generated

Summary

Retrieves a task-matched cognitive scaffold from the Ejentum Reasoning Harness in "memory" mode, which sharpens an existing observation about cross-turn drift, contradiction, or accumulated context so it can be injected into a later LLM step.

Inputs

  • query (required) — Short description of the task or observation you are working on (e.g. a noticed contradiction or drift across turns). Trimmed before sending; an empty or whitespace-only value throws an error for that item. Supports expressions, so it can differ per input item.

Output shape

One output item per input item, with json containing mode (always "memory"), the trimmed query, and response (the parsed JSON body returned by the Ejentum API, i.e. the retrieved harness operation).

Each item makes a separate POST to the credential's base URL (default https://api.ejentum.com/harness/) with body {query, mode: "memory"}, authenticated via the ejentumApi credential. The structure of response is defined by the API, not the node. With Continue On Fail enabled, failed items return json {error: message} instead of stopping the workflow. The node is usable as an AI tool. This is the dynamic memory mode, available on all tiers; Adaptive Memory Harness is the tier-gated variant.

Examples

Example 1: An AI agent notices that a user's stated preferences contradict earlier turns and wants a scaffold before replying.

Operation: Memory Harness; Query: "user's budget changed from $5k to $20k across turns, reconcile contradiction". Pass response to the next LLM node's prompt.

Example 2: Process a batch of conversation summaries, getting a scaffold for each.

Operation: Memory Harness; Query set to expression {{ $json.observation }}; enable Continue On Fail to keep going on errors.

Links

Discussion