Dashboard guide
After npm run ui, Meridian opens at
http://127.0.0.1:3030. This page explains the current UI, workspaces, run persistence, and settings.
Get started (first screen)
The top of the dashboard is a three-step flow — no separate setup tab required:
- Upload resume — PDF, DOCX, TXT, or MD. Parsed locally to fill your profile and search queries.
- Choose your goal — Find jobs, Upskill, or Change career. This switches the workspace below.
- Start analysis — Start job search (label changes per goal), Try demo first, or Open settings for portals, salary guardrails, and LLM.
Click the Meridian logo anytime for product info, documentation links, and contact email.
Settings panel
Open Settings from the header or step 3. It is one panel with three sections — not just filters:
| Section | What you configure |
|---|---|
| Profile | Name, title, skills, target titles, search queries, resume summary, avoid keywords |
| Search & run | Portal (LinkedIn / Naukri / both), run mode, headed browser, location, salary floors, transition fields |
| LLM | Provider, model, API key, webhook, keyword-only mode, test connection |
Click Save in the panel header. Values are stored in src/config.json on your machine.
Workspaces
Three career modes share run history but show different tabs and outputs:
| Goal | Pipeline mode | Main tabs |
|---|---|---|
| Find jobs | search, assist-apply, demo |
Results · Runs · Jobs · Tracker |
| Upskill | next-role |
Overview · Runs · Learning |
| Change career | career-transition |
Overview · Runs · Plan |
Jobs workspace tabs
- Results — live pipeline progress, run summary, and selected run metrics
- Runs — history, run detail, exports, compare two runs, backup/restore bundle
- Jobs — shortlist and market board for the selected run
- Tracker — application status and email apply drafts
Runs: start, refresh, resume, replay
Meridian persists run state on disk so work is not lost casually:
- Page refresh — the agent keeps running; the UI reconnects and shows live progress.
- UI server restart — if the worker is still alive, the server reattaches and tails the log. If the worker stopped mid-run, the run is marked interrupted when a checkpoint exists.
- Resume from checkpoint — on an interrupted run, use Resume from checkpoint to continue from the last completed agent (same run id).
- Replay run — on completed or failed runs, replay with the same mode, portal, and settings as a new run.
Checkpoints are saved after each pipeline agent to data/ui/checkpoints/{runId}.json.
Local data paths
data/output/— jobs, shortlists, tailored resumes, agent reports, strategy JSONdata/ui/run-history.json— run metadata and summariesdata/ui/run-logs/{runId}.log— per-run logsdata/ui/active-run.json— in-progress run snapshot for reconnectdata/ui/checkpoints/— resume checkpoints per rundata/browser-state/— LinkedIn/Naukri login sessions
Documentation ↔ dashboard
This docs site links back to your self-hosted dashboard (header, footer, sidebar). Custom host or port:
?dashboard=https://your-host:3030 on any docs URL — see
Configuration.