Memory workflows cover the experiential side of Spectron: sessions, turns, remember / recall / reflect / forget, profiles, and tuning. Extracted records still live in the same SurrealDB graph as document knowledge – distinguished by provenance and trust (source.kind = "turn", reflections, elaborations, consolidations, …) rather than by a separate database.
The experiential side is where “I have a cat”, “I saw a cat last night”, and “I used to have a cat” land as different kinds of belief – not one undated note. Start with the six experiential categories (including episodic raw turns) in Memory categories, then read Traces and memory evolution to see how response_trace and retrieval_trace close the loop on ranking and reuse.
What this section contains
Sessions – containers for turns; the episodic substrate every extraction cites (creating sessions, adding turns)
Reasoning model – extraction, reconciliation, authority, temporality (extraction pipeline)
Tuning – models per stage, caching, ontology
HTTP verbs (/facts, /facts/batch, /query, /chat, /reflect, traces) are documented under Surface, models, and security and the REST API.
Authoritative ingest is covered in Knowledge.