Skip to content

Add isolated conversational development runtime - #2

Draft
DevHabito wants to merge 4 commits into
codex/v50-kernel-foundationfrom
codex/conversational-darwin-dev
Draft

Add isolated conversational development runtime#2
DevHabito wants to merge 4 commits into
codex/v50-kernel-foundationfrom
codex/conversational-darwin-dev

Conversation

@DevHabito

Copy link
Copy Markdown
Owner

What changed

  • pre-registers E044 before implementation
  • adds an isolated conversational runtime outside the frozen E043 desktop runtime
  • requires explicit backend and model selection with no cross-provider fallback
  • adds an OpenAI Responses adapter with mandatory model probing, store: false, strict Structured Outputs, no tools, and no provider-side continuation id
  • performs model-backed UNDERSTAND followed by deterministic policy processing and model-backed EXPRESS
  • keeps at most 60 temporary messages in memory and clears them on close
  • exposes no persistent-memory, goal, RZS, sigma, identity, world-model, consent, capability, or action mutation handle
  • adds an explicitly supplied local-backend seam without installing or discovering a local server
  • adds a terminal command and an English setup guide

Why

Darwin needs an open-ended conversational development surface without turning the language model into the cognitive authority or contaminating the PURE E043 candidate. The protocol and implementation state explicitly that this remains an LLM-centered adapter, not evidence of consciousness, general cognition, or a Diana-like entity.

Validation

  • E044 protocol committed first: 61d6f97
  • implementation commit: ca04a2c
  • 24/24 focused E044 tests passed
  • full suite: 472 discovered, 471 executed and passed, 1 pre-existing Windows symlink skip, 0 failures
  • E043 runtime SHA-256 unchanged: fd0d8aaf2bd1011addea581eaef172ce940157164dc013681d5326476d49f7e8
  • E043 protocol SHA-256 unchanged: beea70ce6bcfd177a18d36feca016f5f93ed0bed7743fc73c31152cc19eaefad
  • captured request tests verify configured model, store: false, strict schemas, and absence of tools and previous_response_id
  • no live API or local-model call was made

Evidence ceiling

The OpenAI key and an explicit local backend were both unavailable on the test machine. Live model probing, live UNDERSTAND, live EXPRESS, the 20-30 minute conversation probe, language quality, and topic-return quality remain UNEXECUTED or UNKNOWN. E041/E042 remain independently blocked on human annotation.

This is a stacked draft PR targeting codex/v50-kernel-foundation so its diff contains only E044.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant