How does HELLO Cortex stop agents from repeating mistakes?

DIRECT ANSWER

hello.cortex keeps agent memory in three tiers: CORE (working), RECALL (retrievable) and ARCHIVE (cold). Failures are recorded with their fixes, and before retrying, an agent can query for similar past failures using cosine similarity over local vectors. Contract locks record decisions agents agreed on. Automatic sleep-cycle consolidation is a design goal, and no token savings are claimed.

1. Tiered memory

Each memory row has a tier and an agent. Recall queries return the most relevant rows for that agent.

2. Failure registry

POST a failure, then query /api/failures/similar with a new error message to get the closest past failures and how they were fixed. Recurrences are counted.

3. Consolidation

Promoting high-confidence episodic memories into shared skills during idle time is planned (vision).

What exists in the repo

All docs pages