Elamur Groupantsworld-dialogue
The groupant beside models/world-dialogue.groupant in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.
Facts
Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.
Components
| component | label | where |
|---|---|---|
| approximant | partial | forward |
| ast | not materialized | |
| composition | partial | composition |
| cost | symbolic source derived | cost |
| descriptant | present | descriptant |
| invariant | partial | invariant |
| tests | see source local evidence | checks |
Obernants
The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.
| id | type | label | where |
|---|---|---|---|
behavior-complement | behavior_complement | partial source derived model | behavior-complement |
reverse-reconstruction | reverse_reconstruction | partial source derived model | reverse-reconstruction |
normalization-join | representation_transport | partial source derived model | normalization-join |
coordinate-join | coordinate_transport | partial source derived model | coordinate-join |
effect-order | failure_precedence | partial source derived model | effect-order |
resource-join | cost_composition | partial source derived model | resource-join |
Model text
descriptant
world-dialogue: wave28
ideas->nerves::norm, ideas->talk::now, talk->babi::split_sentences, main->vmm->lang/world. Act cue normalization and concept normalization are not equivalent adapters.
This source-bound partial group organizes existing Mova code. Native evidence and excluded effects are recorded in source sidecars and the root wave28 report. Registration is not a universal proof or whole-system runtime claim.
model
world-dialogue
forward
lib exports dialogue/graph/persistence/VMM; main activates state then dispatches. ideas related joins norms from nerves and timestamps from talk. VMM prompts serialize story+world views, teacher text parses into world commands, repairs address physical lines.
invariant
Pure helpers are deterministic under fixed inputs; clock/file/state/teacher paths require fixed external observations. Case/plural projection loses information; repair acceptance is not gate validity.
cost
Argument dispatch plus three sparse diffusion rounds, rarity index and rank sorting, whole-memory sorting; file/teacher/executor costs additive at actual call sites, no uniform speed claim.
composition
ideas->nerves::norm, ideas->talk::now, talk->babi::split_sentences, main->vmm->lang/world. Act cue normalization and concept normalization are not equivalent adapters.
domain
Bounded UTF-8 synthetic inputs and fresh temporary files; finite representable integer intermediates, positive max for nonempty VMM parts and sentence counts within u16 for intended ranges.
limitations
No live API/teacher, DB, corpus, model mutation, GPU or full-school run; lib visibility compile only, main startup excluded.
checks
- wave28_dialogue selected nonempty tests twice debug/release
- Compiled source mutants and restored positive
- Parent syntax generated twice identically
catalog
behavior-complement
Exclude live teacher or state mutation from pure classification; exclude counting duplicate query words twice in diffusion/rarity; done-file errors cannot call teacher. No guarantee that repaired lines pass execution gates.
reverse-reconstruction
Joint observations admit all texts with same normalized concept sets and act priority, files sharing retained ranked rows, and cached histories yielding same command bytes. These fibers are set-valued and do not recover overwritten memories or raw repair history.
normalization-join
ideas words = set(filter_map(nerves::norm, alphanumeric tokens)); talk cue matching has different boundaries and cannot substitute for nerves normalization.
coordinate-join
VMM splice uses 1-based physical lines while prompt partitions use 1-based inclusive sentence anchors; changing line N does not imply changing sentence N.
effect-order
Main activates state before dispatch; VMM done refusal precedes partition and teacher; ideas atomic write failure is swallowed, unlike VMM file errors.
resource-join
Three diffusion frontiers bounded at 4000 coexist with unbounded total activation and whole-file ideas/memory loading; teacher cost has no measured time bound.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/models/world-dialogue.groupant/