Skip to content

feat(modes): add native compact and extreme surfaces - #4

Merged
lennney merged 2 commits into
mainfrom
feature/v2-modes-public-handoff
Aug 9, 2026
Merged

feat(modes): add native compact and extreme surfaces#4
lennney merged 2 commits into
mainfrom
feature/v2-modes-public-handoff

Conversation

@lennney

@lennney lennney commented Aug 8, 2026

Copy link
Copy Markdown
Owner

Summary

  • Introduce the public Native, Compact, and Extreme product modes for the
    0.2.0-alpha.1 line.
  • Preserve complete authorized tool schemas for Compact and Extreme discovery,
    validate arguments against the original schema before contacting upstream,
    and keep exact result recovery through read_result.
  • Replace legacy compressor and model-selection evaluation paths with the
    public mode evaluation runner, schemas, deterministic evidence, and agent
    engineering contract.

Why

The new surface keeps a small, predictable tool interface where it helps Host
compatibility while preserving parameter completeness and recoverable delivery.
Native remains available for Hosts that need the normal authorized catalog.

Validation

  • npm run ci
  • 43 test files passed, 499 tests passed, 1 skipped
  • Coverage: 83.46% statements, 78.12% branches, 93.29% functions
  • Package boundary: 73 files, 403284 unpacked bytes, zero source maps
  • Public-boundary review removed private worktree naming from contribution docs

Safety and scope

  • Configuration is version 2; version 1 is rejected without automatic
    migration.
  • No private workbench plans, hidden tasks, thresholds, credentials, or Host
    model calls are included.
  • This PR does not publish a release or make registry changes.

Copilot AI lite review requested due to automatic review settings August 8, 2026 12:53

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot wasn't able to review this pull request because it exceeds the maximum number of lines (20,000). Try reducing the number of changed lines and requesting a review from Copilot again.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: dcd53747e0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread README.md
Comment on lines +126 to +128
boundary. Extreme uses a smaller boundary only when the initial delivered response
is at least half the size of the exact response; otherwise it passes the exact
response through unchanged.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge State Extreme's 50% threshold in the correct direction

For an Extreme response whose projected delivery is more than half the exact response, the implementation passes it through (minimumSavingsRatio: 0.5 and the check in src/result-capsule-store.ts); projection occurs only when the delivered response is at most half the original size. This text says the opposite, which can mislead users evaluating when Extreme will activate. The Chinese README already describes the intended “at least half smaller” behavior.

Useful? React with 👍 / 👎.

@lennney
lennney merged commit 886352e into main Aug 9, 2026
4 checks passed
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.

2 participants