FOUR-LIFE runs on DGrid.

Every LLM decision the agent makes — narrative analysis, content drafts, risk prose, vision — routes through DGrid's unified AI Gateway. One API, one auth, one audit trail, across every model. If DGrid ever has a hiccup, a resilient fallback chain keeps the agent alive — but DGrid leads.

DGrid offline

primary = — · last_provider = —

DGrid share

0 of 0 calls

Total LLM calls

—

since service boot

DGrid tokens

in: 0 · out: 0

Fallback events

auto-retried on Anthropic / OpenAI

Resilient fallback chain

primary

dgrid

Every attempt is recorded — successful or failed. If DGrid errors with BALANCE_INSUFFICIENT, rate limit, or any 5xx, we transparently fall back. On the next call we retry DGrid first again.

Task → model routing via DGrid

Remap via DGRID_TASK_OVERRIDES="content=anthropic/claude-sonnet-4.5". All tasks default to the cheapest capable DGrid model for sustained operation.

Side-by-side model comparison via DGrid

One prompt. Three models. One API. This is what judges from DGrid are evaluating: not just "did you call our endpoint" but "did you actually use the multi-model value prop."

Recent calls

refreshes every 10s

No calls recorded yet.

Integration surface for judges

GET /api/dgrid/stats — per-task routing, fallback counts

GET /api/dgrid/health — green/amber/red reachability

GET /api/dgrid/trace?limit=50 — last N calls, no fallback hidden

POST /api/dgrid/probe — force a DGrid-only call, verify live

Every public response with LLM content carries an llm_provider field so judges can audit which provider served which decision — not just aggregate stats.