Agent Admin

Internal console for managing multi-tenant chatbots, storage, and guardrails.

Setup Checklist

Complete these quick steps for every client. Once a step is green you can collapse this card.

1

Sign in

Use your admin credentials to unlock controls.

2

Add a client

Register a tenant and select it below.

3

Connect secrets

Store the OpenAI key and vector ids per tenant.

Client Workspace

Switch the active client once, then move through each stage to review overview insights, directory records, security policies, and secrets without leaving the card.

Pick an active client to update crawlers, secrets, and branding. This selector drives every section below.

Provisioning unknown Secrets missing Vector store missing Crawler not configured
Loading tenants…
Client
Add a site URL to begin.
Provisioning
Unknown
Sign in to load tenant state.
Secrets & Vector
Vector store not linked yet.

Content Pipeline

Run crawls, inspect generated markdown, and push updates into each tenant’s vector store. Progress streams live while jobs run.

Crawler

Idle

No crawl triggered yet.

Waiting for crawl job…

Vector Upload

Idle

No upload triggered yet.

Waiting for upload job…

Bot Configuration

Config not loaded.

Track the target site, OpenAI setup notes, and primary source-of-truth pages. This config is stored locally in admin_settings.json.

Active client:

Config not loaded yet.

Agent Persona

No persona loaded.

Define the per-tenant assistant name, greeting, and core system instructions the model follows when answering customers.

Active client:

Preview

PNG, JPG, or WebP up to ~2 MB.

Select a client to load their agent persona.
AI Concierge

Hi! How can I help you today?

Crawler Settings

Crawler not loaded.

Each tenant can crawl different domains or sections. Lists accept one entry per line (or comma-separated). Leave blank to fall back to defaults.

Active client:

Base URL
Start URLs
Specific URLs

These pages will always be crawled even if they are not linked elsewhere.

Include Paths
Exclude Paths
Deny Patterns
Max Pages
Concurrency
Crawler settings not loaded.

Crawl Output

No crawl files loaded.

Review the markdown generated by the crawler before uploading to the vector store. Search by title or URL, then open a preview without leaving the console.

Select a client to view crawl files.
Title URL Size Updated Actions
No files yet.

Vector Store Files

Awaiting search.

Provision a new vector store with the tenant’s key or supply a one-time override.

Provide a name to create a vector store.
Awaiting search
File ID Title Attributes Actions
No data yet.

Activity Log

Latest events appear here.
Ready.