Usage, health & teams

The platform pages cover operations: Usage & Metrics and Health & Latency (/space/governance), Cron (/space/cron), Notifications (/space/notifications), Projects (/space/engineering) and Teams (/space/teams).

New to the AI side? Read AI concepts from scratch first — the metrics on this page (tokens, latency, cost) are exactly the economics of the token-by-token loop.

The concepts behind governance

This page is where the economics of tokens become visible:

• Tokens are the unit you pay for. Every request bills prompt tokens + completion tokens (and cached/reasoning variants where applicable). Nothing on this page is abstract — the numbers come straight from per-request token accounting (usage_records), so “cost” is literally tokens × provider pricing.
• Latency is the cost of reasoning (and of stream length). Higher P95 often tracks reasoning models (which think longer) or longer outputs. Health and the retry handler share the same telemetry, so what you see here matches what the client feels.
• Agents multiply usage. Orchestration runs many steps (orchestrator → executors → verifiers), each writing its own usage record — so a single task can show up as many line-items. That’s the price of composition.

See AI concepts from scratch for why tokens and context drive everything the platform meters.

Usage & metrics (/space/governance)

The governance dashboard (GovernanceDashboard) reports real usage data backed by the usage_records table — every chat/agent/content request writes a record with:

• Token accounting: prompt tokens, completion tokens, cached input, uncached input, reasoning tokens.
• Cost: costUsd computed per request from provider pricing.
• Latency: durationMs.
• Mode: chat / agent / code / workflow / tool.
• Tier: free/paid at request time.

The dashboard aggregates by time range, provider, model and mode: total cost, token volumes, and per-model breakdowns — so you can see exactly what your BYOK usage costs and where it goes.

Health & latency

The Health tab (HealthTab) is real observability per API request:

• P95 latency across providers/models.
• Success rate and error rate per request.
• Risk level per provider request — surfaced from src/lib/request-metrics.ts and the health API.

This is the same telemetry used by the ops panel and the retry handler, so what you see in health matches what the client experiences.

Cron (/space/cron)

Cron (CronManager) schedules autonomous tasks on Cloudflare Workers (src/lib/cron-jobs.ts, cronManager):

• Define a task with a cron expression and an action (e.g. run a workflow, ping a channel, checkpoint a repo).
• Scheduled runs execute at the edge; each run’s result is recorded.
• Combine with the Pipeline Studio to run entire graphs on a schedule instead of manually.

Notifications (/space/notifications)

The notification center (NotificationCenter) collects real-time agent alerts and webhook configuration:

• Alerts from agent runs (completion, failure, approval requests).
• Webhook configuration to forward alerts to external systems.
• Inbound channels (Slack etc.) surface their messages here too.

Projects & repositories (/space/engineering)

The engineering dashboard (EngineeringDashboard) organizes work around projects and repos:

• Projects group objectives; each project can link GitHub repositories.
• PRs and issues are monitored for linked repos.
• Chat sessions associated with a project/repo appear in the session history tagged as “code” sessions.

Teams (/space/teams)

Teams bring real-time collaboration (TeamsPanel, backed by /api/teams):

• Team model: a team has an owner, a slug, members with roles (owner / member), invitation status (pending / accepted) and joinedAt.
• Stats cards: total teams, active members, pending invites, teams you own.
• Collaborative sessions: multiple users in the same thread, code review, and coordinated PR delivery — the same conversation model shared across members.

Settings (/space/settings)

The settings page (DashboardSettings in settings-panel.tsx) covers: profile (name/avatar), BYOK provider keys with the encrypted vault, integration tokens (GitHub, Google Drive BYOS), language, theme, and the cloud-sync opt-ins for vault entries. See Models & API keys for the vault details.