GrayMatterEnvironment

Pre-context + knowledge chunks + flash. Calls neuron_get_context and neurag_query in parallel, merges the results, uses the cache.

Show Gray-Matter status: registered servers, cache, counters.

Persist a cross-store bridge: a link between a Neuron concept and a NeuRAG knowledge node the orchestrator found to relate. Recalled in future pulses on either endpoint.

Blackboard (GM, thalamus): publish key=value with an optional TTL in seconds. Keys are 'org/component/name'. Returns the entry with its version.

Blackboard: read key. None if missing or expired — state decays on its own.

Blackboard: changes from version `since` onwards, filtered by key prefix (expired entries included).

Brain (GM, thalamus): what surrounds a problem, with its history. Mid-band memory nodes (related but not obvious) with their facts and link rationales — a past decision, a bug solved once — plus the nearest knowledge chunks. For dilemmas, problems and decisions. Requires 'seed'.

Graph-ize a folder OR a single document server-side in ONE

Status of ingest jobs started with knowledge_ingest

Chunk a file or directory without saving. Returns JSON list of chunks. LLM then calls knowledge_add_node + knowledge_add_chunks to organize them. For whole folders prefer knowledge_ingest (server-side, no chunks through context).

Create a node in the hierarchy.

Attach previously indexed chunks to a node.

Show knowledge base status: engine, node count, chunk count.

Show the hierarchical node tree.

Structural audit of the vault: broken hierarchy, tiny/empty chunks, duplicate names (serious) + orphan nodes, chunks without source, nodes without triggers (warnings). Read-only — flags, never deletes.

Show all node links (tag_overlap, cross_ref) with weights and evidence.

Clear all links and rebuild from tags + cross-refs. Returns count of links created.

Re-embed every chunk with the currently active embedding model. Use after changing embed_model: vectors from two models are not comparable, so search returns noise until the vault is rebuilt. Only vectors change — chunk text, nodes and links are untouched, and the source files are not needed. For a …

D3 — structured neighborhood of a node, resolved from a query (trigger match, then exact name). BFS over parent/children/links up to `depth` hops. JSON: {node, neighbors:[{name, path, node_type, relation, distance}]}. Empty node = no match. Cheap (SQL-only) — built for Gray Matter's proactive-knowle…

Mark results as having been USEFUL TOGETHER, so the links

Associative expansion: spreading activation from a node, k hops

Delete a node and its entire subtree (children, chunks, links).

Rename a node; updates the materialised path of itself and all

Bulk-import a folder tree from a YAML mapping (deterministic, no

Return the FULL text of a NeuRAG skill on demand — token-cheap,

Screening: find existing similar keywords (vector search). Call BEFORE store_turn.

Semantic vector search. Find similar keywords via Turso vector_distance_cos or a Python cosine fallback (384-dim fastembed embeddings, NS_EMBED_MODEL).

Textual graph summary: top keywords, recent links, health, forgotten concepts

Neuron self-model (C3): what the memory knows about itself — strongest/most-trusted concepts, recent growth, weakest domain, loop compliance. JSON.

Find keywords not touched in N turns (decaying salience). Useful for rediscovering lost concepts. With `near`, ranks dormant concepts by mid-band similarity to a topic — relevant but non-obvious (serendipity).

Force prune inactive tangential links

Toggle keyword deduplication, or set it explicitly with enable=true/false. Output reports the resulting ON/OFF state.

Toggle semantic flashbacks

Reset the graph and start over. DESTRUCTIVE & irreversible: requires confirm=true.

Automatic semantic extraction from text: keyword, topic, domain, intent, sentiment, entities. Heuristic (0 token) — no LLM extraction.

POST fallback (0-token): one-shot extract + topic-shift + auto-link + save. Prefer a curated store_turn when you can pick the concepts yourself; use auto only for throwaway turns.

Export the complete graph as JSON

Switch active context (creates if new). E.g. 'java/spring', 'python/django'.

List all available contexts with metadata.

Show every Neuron command (one line each) plus how to use Neuron well. Call once at the start if unsure; full playbook: call skill(name='playbook').

Return the FULL text of a Neuron skill/playbook on demand — token-cheap, fetch it only when you need the details. Use after the compact opener to load the complete workflow or curation rules.