feat: add long-term memory systems with fused retrieval and memory explorer

Phase 5 (ADR-0003). pkg/concepts fixes the epistemic vocabulary
(observation, fact, memory, claim, hypothesis, belief, prediction,
assumption, inference, opinion), memory kinds and relation names.

internal/memory stores episodic, semantic, procedural and
autobiographical memories in one mind-scoped table with trust-derived
credibility, importance, salience and decaying activation. Retrieval
fuses full-text rank (any-term by default), optional pgvector similarity,
recency, activation, importance and credibility, and returns the
component scores so every recall can be explained. Decay lowers
activation and never deletes; retraction and supersession keep rows for
provenance. Relationships are constrained to one mind at insert time.
Embeddings live in a conditional pgvector table keyed by model so vector
spaces are never mixed; without an embedder the system degrades to text
and graph retrieval.

Attention gates encoding: the attend handler turns admitted observations
into episodic memories (operator changes become autobiographical ones)
with memory_created events; the retrieve process proposes recalled
memories as candidates and admitted recalls are reinforced with
memory_retrieved events carrying their scoring components.

Memory explorer UI (search with component breakdown, timeline, filters,
operator knowledge entry), memory detail with provenance, relationships,
citing events and an inaccurate-memory correction; API; noemactl memory
commands; decay and embed-missing background jobs.

migrations/0005_memory.down.sql

@@ -0,0 +1,3 @@
+DROP TABLE IF EXISTS memory_embeddings;
+DROP TABLE IF EXISTS memory_relationships;
+DROP TABLE IF EXISTS memories;

migrations/0005_memory.up.sql

@@ -0,0 +1,66 @@
+-- Long-term memory: one table, four kinds, explicit epistemic status, decay
+-- without deletion, text search, relationships, and optional pgvector
+-- embeddings (ADR-0003).
+
+CREATE TABLE memories (
+  id                uuid PRIMARY KEY,
+  mind_id           uuid NOT NULL REFERENCES minds(id),
+  kind              text NOT NULL CHECK (kind IN ('episodic','semantic','procedural','autobiographical')),
+  epistemic         text NOT NULL,
+  status            text NOT NULL DEFAULT 'active' CHECK (status IN ('active','retracted','superseded','archived')),
+  content           text NOT NULL,
+  detail            jsonb NOT NULL DEFAULT '{}'::jsonb,
+  source            text NOT NULL DEFAULT '',
+  trust             text NOT NULL DEFAULT 'low',
+  credibility       double precision NOT NULL DEFAULT 0.5,
+  confidence        double precision NOT NULL DEFAULT 0.5 CHECK (confidence >= 0 AND confidence <= 1),
+  importance        double precision NOT NULL DEFAULT 0.5 CHECK (importance >= 0 AND importance <= 1),
+  salience          double precision NOT NULL DEFAULT 0.5,
+  activation        double precision NOT NULL DEFAULT 1,
+  valence           double precision,
+  arousal           double precision,
+  occurred_at       timestamptz NOT NULL DEFAULT now(),
+  created_at        timestamptz NOT NULL DEFAULT now(),
+  updated_at        timestamptz NOT NULL DEFAULT now(),
+  retrieval_count   integer NOT NULL DEFAULT 0,
+  last_retrieved_at timestamptz,
+  origin_event      uuid,
+  supersedes        uuid REFERENCES memories(id),
+  search            tsvector GENERATED ALWAYS AS (to_tsvector('english', content)) STORED
+);
+CREATE INDEX memories_mind_kind_idx ON memories (mind_id, kind, occurred_at DESC);
+CREATE INDEX memories_mind_status_idx ON memories (mind_id, status);
+CREATE INDEX memories_search_idx ON memories USING gin (search);
+CREATE INDEX memories_origin_idx ON memories (origin_event) WHERE origin_event IS NOT NULL;
+
+CREATE TABLE memory_relationships (
+  id          uuid PRIMARY KEY,
+  mind_id     uuid NOT NULL REFERENCES minds(id),
+  from_memory uuid NOT NULL REFERENCES memories(id),
+  to_memory   uuid NOT NULL REFERENCES memories(id),
+  relation    text NOT NULL,
+  confidence  double precision NOT NULL DEFAULT 0.5,
+  source      text NOT NULL DEFAULT '',
+  created_at  timestamptz NOT NULL DEFAULT now(),
+  UNIQUE (from_memory, to_memory, relation)
+);
+CREATE INDEX memory_relationships_to_idx ON memory_relationships (to_memory);
+
+-- Embeddings only when pgvector is present. The vector column is untyped so
+-- several models with different dimensions can coexist; rows carry the model
+-- name and dimension so representations are never mixed.
+DO $$
+BEGIN
+  IF EXISTS (SELECT 1 FROM pg_available_extensions WHERE name = 'vector') THEN
+    CREATE EXTENSION IF NOT EXISTS vector;
+    CREATE TABLE memory_embeddings (
+      memory_id  uuid NOT NULL REFERENCES memories(id),
+      model      text NOT NULL,
+      dims       integer NOT NULL,
+      embedding  vector NOT NULL,
+      created_at timestamptz NOT NULL DEFAULT now(),
+      PRIMARY KEY (memory_id, model)
+    );
+    CREATE INDEX memory_embeddings_model_idx ON memory_embeddings (model);
+  END IF;
+END $$;