Skip to content

Yuhi v0.2.0 Beta 2

Pre-release
Pre-release

Choose a tag to compare

View all tags
@YUHI-AI-Labs YUHI-AI-Labs released this 29 Jul 01:16
v0.2.0-beta.2
51be0c1

Yuhi 0.2.0 beta — release notes (draft)

  • CLI (npm): @yuhi-ai-labs/yuhi@0.2.0-beta.1
  • VS Code (Marketplace): yuhi-vscode 0.2.0 — publish as pre-release (vsce publish --pre-release)
  • Coherent "0.2.0 beta" line. Root monorepo stays private; internal @yuhi/* packages are not published.

What's available

  • CLI preview: yuhi preview / status / explain / diff / scan / doctor.
  • Local context preparation (yuhi prepare .): deterministic scan → local summarization
    (Ollama) → pseudonymization → deterministic safety-check → inspectable prepared workspace.
  • Secret scanning + pseudonymization (names/IDs) with masking.
  • Generated prepared workspace under .yuhi/prepared/<id>/ — source files never modified.
  • Context Savings led by Estimated Claude input avoided.
  • VS Code beta (Doctor / Setup Local AI / Prepare Workspace / Review + Explorer "Prepare with Yuhi",
    Original↔Prepared diff, status states incl. inference-verified readiness) — installable, pending
    Marketplace publication     (do not describe as "Available" until published).

Limitations (honest)

  • Token counts are estimates (chars/4 heuristic; a model tokenizer can be injected later).
  • Actual AI usage and billing vary by product, provider behavior, caching, and pricing — no guaranteed savings.
  • Local summarization requires Ollama + an installed model (qwen3:1.7b default; qwen3:4b quality option).
  • Local-model quality/latency depend on hardware; older Ollama versions may fail to run newer models (update Ollama).
  • No automatic forwarding to Claude/Codex — human review is required and recommended.
  • GUI acceptance (clean-profile click-through) requires a final manual pass.

Install

# CLI (after npm publish)
npx @yuhi-ai-labs/yuhi@latest preview
npm i -g @yuhi-ai-labs/yuhi

# Local AI
# install Ollama from https://ollama.com/download, then:
yuhi setup-local-ai      # pulls qwen3:1.7b after explicit confirmation
yuhi prepare .

Artifacts + checksums

Built under /tmp/yuhi-release-work/artifacts/ (not committed):

  • yuhi-ai-labs-yuhi-0.2.0-beta.1.tgz
  • yuhi-vscode-0.2.0.vsix
  • SHA256SUMS.txt

Handoff commands — DO NOT RUN without explicit approval (irreversible/remote)

# 1) Clean-profile VS Code GUI test (yours)
TMP=$(mktemp -d)
code --user-data-dir "$TMP/user" --extensions-dir "$TMP/ext" --install-extension apps/vscode/yuhi.vsix
code --user-data-dir "$TMP/user" --extensions-dir "$TMP/ext" examples/acceptance-fixture

# 2) npm publish (CLI)
cd apps/cli && npm publish --access public   # 2FA/OTP or automation token required

# 3) VS Code Marketplace (pre-release)
cd apps/vscode
npx vsce login yuhi-ai-labs                   # confirm publisher identity (permanent)
npx vsce publish --pre-release --packagePath yuhi.vsix

# 4) GitHub push + tag + release (from repo root)
git push origin main
git tag v0.2.0-beta.1 && git push origin v0.2.0-beta.1
gh release create v0.2.0-beta.1 --prerelease --title "Yuhi 0.2.0 beta" --notes-file docs/RELEASE_NOTES_0.2.0-beta.md

# 5) Post-publication verification
npx --yes @yuhi-ai-labs/yuhi@latest --version
mkdir -p /tmp/yuhi-live && cd /tmp/yuhi-live && npx --yes @yuhi-ai-labs/yuhi@latest preview

After publishing: rotate any npm/Marketplace tokens; update the site "VS Code" status to
Available — Beta and add the Marketplace install button only once the live URL exists.

Assets 5
Loading