auto_scholar

auto_scholar is a workspace for paper-oriented research automation.

v0.1 scope

This release is a local-first, zero-install scaffold for reading and triaging generated paper notes. It is suitable for personal use while the backend/API contract is still forming.

Included surfaces:

• repo-local skills for generating Chinese-first paper articles and cleaning paper tags
• Web UI paper tracking with read/recommend/archive state
• Markdown reader for article.md packages under the configured knowledge base
• local image serving for article assets under each paper package's assets/ directory

Out of scope for v0.1:

• multi-user sync
• writing newly collected Web UI intake papers back to the repository
• authentication or remote deployment

Configuration

Local repository configuration lives in .as/config.yaml, which is intentionally ignored by git.

knowledge_base:
  personal_path: ../my_autoscholar

The Web UI reads knowledge_base.personal_path as the Markdown paper library root.

Web UI

The first frontend scaffold lives in webui/. It is intentionally dependency-light so the repo has a usable interface before the backend contract is finalized.

Current surfaces:

• paper tracking: status, priority, unread updates, filters
• paper intake: collect a new paper into a collection with seed metadata
• markdown reader: render structured paper notes, claims, tables, code blocks, and timeline

Run locally:

cd webui
node scripts/serve.mjs

Then open http://127.0.0.1:5173.

Validate the static scaffold:

cd webui
node scripts/check-js.mjs
node scripts/smoke-test.mjs

If npm is available, npm test runs the same checks. The smoke test covers the Markdown renderer, article metadata parsing, asset path rewriting, and the local server's config/paper APIs.