Document projector seam before compaction

This commit is contained in:
Meta Vens 2026-05-02 03:32:37 +01:00
commit 92df5142cb
4 changed files with 32 additions and 12 deletions

View file

@ -118,9 +118,9 @@ training feedstock for projection and dialogue scaffolding.
## Current Next Action
Materialize the accepted v5 thinking-plus-tools sequential capture into an Ink
branch draft and sidecar, then apply reviewed state mutation only after
validation.
Build the projector artifact seam that translates canonical state into compact
character-local operating context, then rerun or materialize v5-style
sequential generation through that projector rather than raw state soup.
Cat/Oz remains useful as an Elysium procedural mechanics fixture, but grounded
training data should start from authored historical lore rather than gameplay
@ -196,13 +196,22 @@ Completed projection path items:
v3 preserved a useful failure where old prompt text still caused plain JSON
content in later passes; v5 is accepted as draft and produced valid
tool-shaped Maer choice, Sella appraisal, and Sella next action
- identify the projector gap:
the sequential runner is still scaffolding because it sends selected raw
numeric state variables into prompts. The next real organ is a projector or
state interpreter that turns canonical variables, memories, relationship
stance, culture, and scene pressure into readable character-local operating
context and action affordances. Character agents should not need to interpret
`current_activation`, `plasticity`, or numeric dimension scores.
Remaining projection path:
- build a deterministic speaker-local input slicer
- build a renderer that emits projection controls before prompt prose
- materialize the accepted v5 thinking-plus-tools sequential capture into Ink
and sidecar annotation
- build the projector artifact seam before materializing more generated
branches
- rerun or adapt v5-style thinking-plus-tools sequential generation through
projected character-local context instead of raw state variables
- improve the local-awareness-to-Ink prompt so semantic `training_hooks` do not
drift into future branch ids, object custody remains branch-local, and action
labels come from the exact canonical enum

View file

@ -209,6 +209,7 @@ Ink is now the standard branching scene format. The first Qwen-generated Ink
draft now exists with a reviewed capture, the first selected branch now has a
reviewed mutation replay that updates both involved characters, v2 shows the
symmetrical turn model improves action behavior, and v5 proves local
thinking-plus-tools invocation works for the scene runner. The next
implementation target is materializing the accepted v5 capture into Ink and a
sidecar annotation.
thinking-plus-tools invocation works for the scene runner. The runner is still
scaffolding because it sends selected raw numeric state variables into the
model. The next implementation target is the projector artifact seam: canonical
state becomes compact character-local operating context before Qwen sees it.

View file

@ -114,8 +114,15 @@
text conflicted with tool calling.
- Fifth sequential draft: v5 is accepted as draft. It used thinking plus
native tools for Maer choice, Sella appraisal, and Sella next action.
- Next prototype: materialize the accepted v5 capture into Ink and sidecar
annotation, then apply reviewed state mutation only after validation.
- Projector gap: the sequential runner still sends selected raw numeric
state variables into prompts. This is acceptable scaffolding for Qwen
plumbing tests, not the final Ghostlight architecture.
- Next prototype: build the projector artifact seam that turns canonical
variables, memory, relationship stance, culture, and scene pressure into
compact character-local operating context and action affordances.
- After projector seam exists, rerun or adapt v5-style thinking-plus-tools
sequential generation through projected context, then materialize only a
validated branch into Ink and sidecar annotation.
5. Build the projection distillation loop. Started.
- Use `docs/architecture/projection-distillation-plan.md` as the teacher to

View file

@ -118,7 +118,10 @@ accepted_design:
- The second sequential Qwen capture validates the symmetrical turn shape behaviorally by producing canonical action labels and a non-speech Sella action, but still needs revision because prose prompting did not enforce array-typed JSON fields.
- Qwen structured generation should use Ollama `/api/chat` with native tools and `think: true`, not `/api/generate` plus prose-only JSON instructions; Qwen's XML-shaped tool format is an internal chat-template/tool-call convention, not a reason to wrap Ghostlight artifacts in XML.
- The local Qwen tool path works with thinking enabled, but nested arguments can still double-stringify, so the runner repairs known nested fields and preserves failures as reviewed captures.
- The current sequential Qwen runner is scaffolding, not the final projection architecture: it still sends selected raw numeric state variables directly to the model.
- A projector/state interpreter must sit between canonical agent state and Qwen prompts, translating numbers, memories, relationships, culture, and scene pressure into compact character-local operating context and affordances.
- Character agents should not need to interpret `current_activation`, `plasticity`, or raw dimension scores; those belong to Ghostlight's internal state model and projector training data.
current_status:
summary: Ghostlight now has the Epiphany-style persistence spine, the first architecture payload, the v0 canonical agent-state schema seam, prompt projection docs, immediate schema fixes from evaluation, first reviewed Cat/Oz projection examples, a revised Aetheria lore-grounding architecture, a lore grounding digest schema, the first draft Cold Wake historical grounding digest, a Cold Wake story-lab fixture, projection controls, validated Qwen capture receipts, a complete Cold Wake short story assembled from accepted projection outputs, a hand-authored Ink-backed Cold Wake branching scene, the first Qwen-generated Ink draft materialized from local awareness with reviewed training annotations, the first reviewed sequential branch mutation that updates both Maer and Sella after a sanctuary intake beat, a v2 sequential Qwen capture proving the symmetrical turn shape produces better action behavior, and a v5 sequential Qwen capture proving local thinking-plus-tools invocation works for the scene runner.
next_action: Materialize the accepted v5 thinking-plus-tools sequential capture into an Ink branch draft and sidecar, then apply reviewed state mutation only after validation.
summary: Ghostlight now has the Epiphany-style persistence spine, the first architecture payload, the v0 canonical agent-state schema seam, prompt projection docs, immediate schema fixes from evaluation, first reviewed Cat/Oz projection examples, a revised Aetheria lore-grounding architecture, a lore grounding digest schema, the first draft Cold Wake historical grounding digest, a Cold Wake story-lab fixture, projection controls, validated Qwen capture receipts, a complete Cold Wake short story assembled from accepted projection outputs, a hand-authored Ink-backed Cold Wake branching scene, the first Qwen-generated Ink draft materialized from local awareness with reviewed training annotations, the first reviewed sequential branch mutation that updates both Maer and Sella after a sanctuary intake beat, a v2 sequential Qwen capture proving the symmetrical turn shape produces better action behavior, a v5 sequential Qwen capture proving local thinking-plus-tools invocation works for the scene runner, and a newly explicit projector gap warning that raw state variables must not be the final prompt surface.
next_action: Build the projector artifact seam that translates canonical state into compact character-local operating context, then rerun or materialize v5-style sequential generation through that projector rather than raw state soup.