Architecture

Meridian orchestrates specialized agents locally. The dashboard at src/ui/server.js starts runs and surfaces artifacts from data/output/.

UI server and persistence

src/ui/server.js spawns pipeline workers, streams logs, and writes run state under data/ui/:

  • active-run.json — reconnect after page refresh or server restart
  • run-history.json — completed, failed, and interrupted runs
  • checkpoints/{runId}.json — agent progress for resume
  • run-logs/{runId}.log — stdout/stderr from each run

The pipeline orchestrator saves a checkpoint after each agent. Resume skips completed agents and hydrates state from the checkpoint file.

Job search pipeline

  1. Research — scrape LinkedIn/Naukri via Playwright
  2. Business research — company stability signals
  3. Salary upgrade — compensation benchmarking
  4. JD comparison — fit scoring and shortlist
  5. Resume modifier — per-role tailored resumes
  6. Learning prep — interview prep tasks
  7. Apply agents — assisted Easy Apply / email drafts

Other modes

  • next-role — upskilling and learning paths
  • career-transition — bridge roles and transition strategy
  • demo — sample output without portal access

Extending Meridian

See CONTRIBUTING.md for adding agents and portals.

Dashboard guide — workspaces, settings panel, resume/replay.