Skip to content

Dashboard Usage

Scope: standalone Python Runtime. For Desktop operations, see the Desktop guide.

The Echo Agent Dashboard is a web-based management interface for monitoring system status, managing sessions and resources, viewing logs, and analyzing usage data.

The Dashboard is served by the gateway process rather than on a port of its own; both its address and its authentication come from the gateway configuration.

Architecture Overview

The Dashboard uses a frontend-backend separation architecture:

  • Frontend: Single Page Application (SPA) providing real-time status panels and operation interfaces
  • Backend API: Located in the gateway/api/ directory, responsible for data aggregation and permission validation
  • WebSocket: Used for real-time streaming of logs, session states, and other live data

Accessing the Dashboard

Start the gateway (echo-agent gateway), then open the gateway address in a browser:

http://127.0.0.1:58123/

The Dashboard is a single-page application mounted at the gateway root. Its port is gateway.port (58123 by default; set it to 0 for a dynamically assigned port, which is written to workspace/.echo-agent/gateway.json). When the frontend bundle has not been built, that address falls back to the built-in playground page — run echo-agent dashboard build to produce the bundle.

The simplified page is also available at /playground. It discovers API and interactive WS paths through /meta. Enter an existing API/admin token when authentication is enabled; the page sends it in headers or the WS auth frame, never in a URL or browser storage. Release wheels include the Dashboard and need no Node/pnpm at runtime; those tools are needed to build from source.

The left sidebar contains the main navigation with the following pages:

Page Function
Overview System overview and health status
Sessions Session management
Memory Memory storage browsing
Skills Skills management
Knowledge Knowledge base management
Cron Scheduled job management
Kanban Task board visualization
Logs System logs
Analytics Usage statistics
Config Runtime configuration
Channels Channel configuration

Page Details

Overview

Displays the overall system health dashboard.

Key Features:

  • Active session count
  • Memory entries, running tasks and interactive WS clients
  • Channel connection status
  • Event-driven refresh with polling as a reconnect backstop

Sessions

View and manage all active and historical sessions.

Key Features:

  • View active session list and details
  • Browse historical session records
  • Reset a conversation; page through history and inspect execution records
  • Search session identifiers

Memory

Browse and search memory entries across all tiers.

Key Features:

  • Filter by the four tiers: working, episodic, semantic, archival
  • Search memory content
  • View memory metadata
  • Delete memory entries (the page offers no manual create action)

This page is a cross-subject view: both the listing and the search require an admin token, since a regular API token can only read memories within its own scope.

Skills

Manage installed skills and their configurations.

Key Features:

  • View installed skills list
  • Enable/disable individual skills
  • View skill trigger conditions and descriptions
  • Read skill content, files and dependency status, and install declared dependencies

Knowledge

Manage documents and knowledge base content.

Key Features:

  • Upload documents and trigger an index rebuild
  • Browse indexed documents (path, size, modification time)
  • View index status: document count, chunk count, staleness, last rebuild time

Upload, delete and rebuild are all guarded server-side by the admin guard. The page has no document category or tag features.

Cron

Configure and monitor scheduled tasks.

Key Features:

  • Create/edit/delete scheduled jobs
  • View cron expressions and next execution times
  • Manually trigger job execution
  • View execution history and results

Kanban

Visualize and manage tasks using a board layout.

Key Features:

  • Drag and drop task cards to change status
  • Toggle visibility of the archived columns (cancelled, suspended)

Dragging requires write access: under a read-only token the board renders read-only, rather than offering controls that answer 403 on every drag.

Logs

View system logs in real time.

Key Features:

  • Real-time log streaming (via WebSocket)
  • Filter by level (DEBUG / INFO / WARN / ERROR)
  • Search text and page through older records
  • Pause or resume live refresh

Analytics

View usage statistics and model invocation costs.

Key Features:

  • Token consumption and daily cost trends
  • Total cost and channel attribution
  • Usage distribution by channel and skill
  • Today, last 7 days and last 30 days

Config

Validate and save editable non-secret settings. An ordinary save does not mutate an already constructed Runtime.

Key Features:

  • Edit language, logging/tracing, memory/knowledge switches and budgets
  • Inspect saved values and redacted running values under _meta.running
  • Restore pending state on refresh or re-login from server-side _meta.pending_paths and _meta.restart_required

Model credentials and gateway authentication remain read-only here; use the server configuration file or supported setup commands. Nested PATCHes merge leaves and cannot bypass secret restrictions. Unknown fields return 400. Saving uses atomic replacement; an invalid existing file returns 409 rather than being overwritten.

Restart reloads the configuration file and environment overrides. Environment precedence may keep a running value different from the file. There is no configuration history/automatic rollback button: restore a backed-up file, validate it and restart. Channel lifecycle operations below are the supported live activation boundary.

Channels

Configure messaging channel connection parameters.

Key Features:

  • View saved enablement, actual connection state and unapplied changes
  • Enable/disable, start/stop/restart configured channels; an explicit lifecycle operation loads that channel's saved configuration
  • Inspect failed or uncertain deliveries; check the destination chat before resuming sends or recording delivery

Maintain channel credentials in the server configuration file; this page does not create accounts or edit credentials.

Access Control

The Dashboard inherits the gateway's authentication; it has no authentication settings of its own. gateway.host defaults to 127.0.0.1, so only the local machine is served.

Management APIs retain API/admin token scopes. Session/turn transcripts require admin scope, including the Dashboard WS sessions subscription. Read tokens receive memory invalidations without entry IDs or principal scopes for aggregate refreshes. Platform/user allowlists separately authorize incoming messages. Same-origin browsers and same-origin WS handshakes without Sec-Fetch-Site work; cross-origin and unexpected Host requests are rejected. Configure allowed_hosts for the reverse proxy and explicitly allow a development frontend's origin only when needed.

Before exposing it publicly

Configure authentication and TLS termination at a reverse proxy before changing gateway.host to a non-loopback address. See gateway authentication and security hardening.

Monitoring and Alerts

Through the Overview and Analytics pages, you can:

  • Monitor system health status in real time
  • Monitor model cost budgets
  • Investigate failures using session/turn state and logs; the page does not configure general resource alerts