Runwise v0.2.0-preview.0

Runwise v0.2.0-preview.0

This preview starts the v0.2 product-iteration line for Runwise.

The focus is the local dashboard: making it wider, clearer, easier to scan, and more useful when reviewing an AI Agent / MCP / RAG project before it goes live.

Runwise is still source-install only. It is not published to npm yet. This is a preview, not a stable release guarantee.

What changed

Redesigned local dashboard

• Wider dashboard layout
• Stronger score and readiness summary
• First-action panel for what to check first
• Severity overview cards
• Findings grouped by severity
• Detected ecosystem cards with local heuristic disclaimer
• Report files section
• Trace / replay / eval next-step hints
• Clear local-first footer

Better dashboard resilience

• Clearer empty states
• Friendlier malformed-report state
• No raw stack traces or [object Object] output

Product iteration foundation

• New version strategy: v0.2, v0.3, v0.4, v0.5 preview lines
• v0.2 focuses on dashboard quality
• v0.3 is planned for Doctor rule depth and false positive feedback
• v0.4 is planned for ecosystem detection expansion
• v0.5 is planned for local JSON rule packs if approved

Verified

• runwise doctor passed
• local dashboard smoke test passed
• trace validate / replay / eval generate passed
• GitHub Actions passed
• .runwise/ outputs remain ignored and untracked

Still not included

• npm package
• GitHub Marketplace release
• hosted service
• telemetry
• login
• model calls
• plugin runtime
• official ecosystem integrations

Runwise v0.1.1-preview.0

Runwise v0.1.1-preview.0

This preview focuses on first-time developer experience, report readability, and release trust.

Runwise is still source-install only. It is not published to npm yet.

What Changed

Easier First Run

• Clean install checklist
• First-run walkthrough
• Test-on-your-project guide
• Clearer pnpm/Corepack guidance
• Clean-machine install review notes

Better Examples

• Example Gallery
• Visual example cards
• Curated Doctor, Trace Replay, Failure-to-Eval, and ecosystem detection samples

Better Reports

• Improved static HTML report hierarchy
• Improved Markdown report readability
• Improved local dashboard readability
• Report reading guide

Clearer Positioning

• Simpler README opening
• Comparison docs
• Less buzzword-heavy GitHub presentation
• Clearer public-preview and source-install boundaries

Future Architecture

• Plugin architecture exploration
• Draft plugin manifest and rule-pack docs
• Documentation-only plugin examples

Plugin support is not implemented yet. These docs explore a possible future direction.

Verified

• Clean clone install passed
• runwise doctor passed
• local viewer smoke test passed
• trace validate / replay / eval generate passed
• GitHub Actions passed
• .runwise/ generated outputs remain ignored and untracked

Still Not Included

• npm package
• plugin runtime
• hosted service
• telemetry
• database
• login
• GitHub Marketplace listing
• stable release guarantee

Try From Source

git clone https://github.com/darwinx687-afk/runwise.git
cd runwise
npx -y pnpm@9.15.4 install --frozen-lockfile
npx -y pnpm@9.15.4 check
npx -y pnpm@9.15.4 test
npx -y pnpm@9.15.4 exec runwise doctor

Runwise v0.1.0-preview.0

Runwise v0.1.0-preview.0

Public Preview Status

This is the first public preview release of Runwise. It is intended for source-based testing, feedback, and early evaluation of the local-first AgentOps workflow.

• Release URL: https://github.com/darwinx687-afk/runwise/releases/tag/v0.1.0-preview.0
• Tag: v0.1.0-preview.0
• Distribution: source install only

Summary

Runwise is a local-first readiness, trace replay, and eval generation toolkit for AI agents, MCP servers, RAG systems, and LLM applications. This prerelease focuses on reviewable local evidence before demos, CI gates, and early release reviews.

Included

• Local-first Runwise Doctor readiness scanner.
• Rule-based readiness engine and readiness score.
• JSON, Markdown, and static HTML Doctor reports.
• Local Dashboard Viewer for generated report JSON.
• Composite GitHub Action readiness gate.
• Local runwise.agent_trace validation.
• Static trace replay reports.
• Deterministic Failure-to-Eval generation.
• Local ecosystem compatibility detection.
• Lightweight examples and trace fixtures.
• English and Simplified Chinese documentation.

Not Included Yet

• npm package publishing.
• GitHub Marketplace listing.
• Hosted service, login, billing, database, or cloud sync.
• Agent runtime orchestration.
• Model calls, model judging, eval execution, or model training.
• Official ecosystem partnerships or official integration claims.

Install From Source

git clone https://github.com/darwinx687-afk/runwise.git
cd runwise
pnpm install

Core Commands

pnpm check
pnpm check:types
pnpm test
pnpm exec runwise doctor
pnpm exec runwise view
pnpm exec runwise trace validate examples/traces/valid-agent-run.json
pnpm exec runwise trace replay examples/traces/mcp-risk-agent-run.json
pnpm exec runwise eval generate examples/traces/mcp-risk-agent-run.json

Known Limitations

• This is a prerelease public preview, not a stable release.
• Distribution is source-based; npm publishing remains deferred.
• GitHub Marketplace release remains deferred.
• Generated .runwise/ artifacts are local, ignored by default, and reproducible.
• The GitHub Action is available from this repository; public versioned action usage should be validated separately before broader adoption.
• GitHub Actions currently reports a non-blocking Node.js 20 deprecation annotation from referenced actions, while CI passes.

Feedback Welcome

Feedback is especially useful on AI agent readiness rules, trace replay readability, Failure-to-Eval output shape, ecosystem compatibility detection, and China-ready LLM provider signals.