Janore
Start free
← Back to docs

Dashboard tour

Your dashboard, explained.

Every tab, what it does, and where to look when something feels off.

Top bar — brand switcher + locale

The brand switcher in the top-left lets you flip between brands (Pro+ tiers). Single-brand accounts see only one entry. The language selector mirrors the dashboard chrome only — your assistants always reply in the visitor's language regardless of what you pick here.

Assistants

The list of assistants you've created. Each card shows status (Live · Paused · Archived), last activity, and a copy-id button. Click into one to see five tabs:

Overview

Embed snippet, channels, languages, persona at a glance. Lifecycle controls live in the header — pause, archive, delete (with a 30-day grace). Status pill shows "X days left" when archived.

Persona

Voice, system prompt, escalation email, primary + secondary languages, LLM provider/model. The system-prompt dropdown lets you start from one of eight starters and tune from there. Use Clear to revert to the default tone derived from the voice you picked.

Knowledge

Two sub-tabs: Sources (URLs, FAQ blocks, manual notes) and Chunks (the actual paragraph-sized rows Janore retrieves at chat time). Inline edit on chunks — saving re-queues them for embedding. Re-crawl on URL sources wipes existing chunks and rebuilds from the latest content.

Channels

Connect WhatsApp / Telegram / Instagram. Web is always on. Each channel shows a status dot (green = receiving messages, amber = setup pending, red = error). One-click "Test" sends a synthetic message and shows the reply.

Test

A live playground that hits the same chat pipeline real visitors hit — same n8n workflow, same vector retrieval, same LLM. Use it to QA copy changes before pushing them to your site.

Knowledge — sources vs chunks

Sources are the inputs you control: URLs, FAQ blocks, free-form notes. Each source row shows a status badge:

  • QueuedCreated, waiting for the next ingest pass.
  • ProcessingCrawling and embedding right now (URL sources can take a minute or two for multi-page sites).
  • IngestedReady. Chunks are searchable.
  • FailedHover the badge for the error message. Common causes: 404, robots.txt block, paywalled page.

Chunks are what the chat pipeline actually searches. ~800 tokens each, embedded into a 1536-dim vector. Editing a chunk resets its embedding so the 5-minute backfill cron re-embeds it. Deleting a source cascades — its chunks vanish too.

Conversations

Every chat across every channel lands here. Filter by assistant or brand. Each row shows: started at, channel, first user message, detected language, last activity, and an Escalated pill if Janore handed off to a human.

Click into a row for the full transcript. Use Mark as escalated for a manual handoff override; this also fires the escalation email if one is configured on the assistant.

Auto-escalation fires when the LLM tags a turn escalate=true (it can't answer, complaint detected, custom-deal request) OR when the lead score crosses 80/100. Either way, an email lands in the assistant's escalation_email within ~30 seconds.

Channels

Top-level view across all your assistants — useful when you have a few brands and want to spot a broken integration fast. Per-channel setup details:

  • Web — paste a script tag, done.
  • WhatsApp — connect a Meta Business phone number.
  • Telegram — paste a bot token from @BotFather.
  • Instagram — link an Instagram Business account.
  • REST API — bearer-auth JSON for custom backends.
  • Generic webhook — Zapier, Make, n8n, anything that can POST JSON.

Billing

Plan, billing cycle, trial expiry, next charge, payment method — all read-only here; changes happen in the Stripe Customer Portal (button at the top of the page). Plan limits: brands, assistants per brand, conversations per month, knowledge chunks. Upgrades take effect immediately; downgrades at end of billing period.

Settings

Profile, notification preferences, diagnostics, danger zone.

  • Diagnostics — one-click probes for the n8n connection, the chat pipeline, and Resend (welcome email). Use these first when something feels broken.
  • Notifications — control which events you get emailed about (escalations, weekly summary, plan limits).
  • Danger zone — delete account (GDPR-compliant 30-day soft-delete, then irreversible).

Pause, archive, delete

Three states for every assistant, each on the Overview tab header:

  • Pause — stops answering visitors immediately. Polite "temporarily unavailable" reply in the visitor's language. No LLM cost, knowledge stays. Resume anytime.
  • Archive — same as pause, plus a 30-day countdown to permanent deletion. Restorable any time before the timer expires. The Overview pill shows "X days left".
  • Delete now — only available on archived assistants. Wipes knowledge, conversations, and channel credentials. Cannot be undone.

When something feels broken

  1. Chat is slow / silent — Settings → Diagnostics → "Test n8n connection". A green pill means the workflow is healthy.
  2. Wrong answers — open the transcript under Conversations. If Janore cited a stale chunk, head to Knowledge → Chunks and edit it inline. Save re-embeds within five minutes.
  3. Escalation emails not arriving — Settings → Notifications → "Send test email" ensures Resend is wired. Check the assistant's Persona tab — escalation_email must be set.
  4. Embed snippet not rendering — see the web embed troubleshooting section.

Where to next?