[Doc]: Make the first successful decision easy to reproduce — community help wanted #86

Motivation

The root README's quickstart builds the Rust frontend, then asks readers to start a model worker using a separate recipe. New users need one complete path from installation to a real decision, with prerequisites and expected output visible up front.

The first page should also explain the practical benefit: fast decision-model serving for agents, followed by a runnable example and a scoped performance result.

Proposed scope

Choose one existing worker as the recommended entry point and document why it is appropriate. A CPU LAYA path is an accessible option; native Open-Jev can be the accelerated path for users with sufficient CUDA hardware. Use #83 and the linked recipes to verify current support.

Start with documentation and existing commands. Add a small launcher only if a reproduction identifies concrete friction that documentation cannot resolve.

Done when

• The entry point states required hardware, RAM/storage, toolchains, dependency versions and model downloads before the commands.
• One contiguous walkthrough covers setup, worker startup, frontend startup, readiness, a request and expected response.
• A contributor unfamiliar with the setup reproduces it and reports the host, repository/model revisions, commands and any blockers.
• Download/export/build time is distinguished from inference latency; no unmeasured “five-minute setup” promise is introduced.
• The README leads with a concrete use case and links directly to the walkthrough, demo and supported-model matrix.
• A concise Chinese entry page carries the same commands, capability limits and links, with a clear maintenance relationship to the English version.
• Stale onboarding statements in touched docs/templates are reconciled with implemented support, and links resolve.

Related: #45 created the documentation site; this issue improves the first-use content. #61 owns shared API contract decisions. #83 tracks model support.

Community help wanted

We need first-time testers, technical writers, translators and contributors who can simplify the walkthrough. CPU-only reproduction and documentation work are welcome.

Please comment with the path you want to test or improve and your environment. A report of the first confusing or failing step is already a useful contribution.