Plan

Gordon is currently in v7 — a split topology of five Rust services, six kellnr-published library crates, a Next.js console, a Python research lab, and a one-shot migrator. The v7 stack is deployed to srv-apps (Pi5, ARM64). Bricks 0–3 are live. Story 24 (staged cutover with live bots) gates production trading.

See Version History for the full v0–v7 arc. See Vision for the North Star.

Phase progress

Phase	Scope	Progress
Phase 2 — Data	gordon-data sole writer, lab read-only, backfill owned	16/16 done
Phase 3 — Executor	NATS intents channel, reconcile-on-restart	10/10 done
Phase 4 — Risk	Five circuit breakers, halt-latch, risk commands on NATS	7/8 done
Phase 5 — Manager	BFF, bot lifecycle, backtest replay, WS fanout	15/17 done
Phase 6 — Bot	Pure strategy engine, NATS klines consumer	11/13 done
Phase 7 — Cutover	Staged rollout to production bots	In flight

Active epics

workspace — cross-cutting

Story	Priority	Status	Notes
DP-04 — Prometheus role + manager-driven target discovery	P0	Done (shipped Wave 2 sub-2b)	~80+ metric families now scraped; alert rules wired
DP-05 — remediation URL chain end-to-end	P0	Done (shipped Wave 2 sub-2a)	138 ErrorCodes × anchor integrity gate
DP-06 — observability baseline (bus tracing + ban-raw-tracing gate + TraceId newtype)	P0	Done (shipped Wave 2 sub-2b)
DP-12+13 — manager /risk/resume proxy + canonical BFF rule	P1	Backend done; console TS finish pending	riskWriterClient console removal is the open tail
cutover — srv-apps v7 deploy	P0	In Progress	Bricks 0–3 live on srv-apps
cutover — staged rollout	P1	Todo	Depends on srv-apps deploy soak
r-06 — testnet 30-day forward test	P1	Blocked (David-gated)	Sequenced after staged cutover + pre-testnet-correctness-gates
op-14 — 24h testnet log-review drill	P0	Todo	Human-gated capstone after r-06
gordon-data auto-precompute on 1m close	P0	Todo	Live WS writes 1m only; bot warmup needs higher TFs
node-20 action deprecation	P1	Todo	Hard deadline 2026-06-02

backbone-simplification (Wave 3b in flight)

39 expert reports across 13 modules. Waves 1, 2, 2.5, 3a shipped. Wave 3b in progress:

• DP-02 position() fix
• DP-03 DataIngestGapDetected wiring
• DP-10 bot_id Prometheus labels
• DP-12+13 BFF rule — manager side done, console TS pending

Net code change through Wave 2.5: approximately -800 LoC across kernel + domain + protocol (deletions outweighed additions).

observability-console-audit (P0)

Console observability gaps identified 2026-05-17:

• DP-07 typed BreakerPayload is half-shipped — manager discards the typed payload and rebuilds from SQL hydration
• DP-16 breaker scope + severity dropped at the WS edge
• DP-06 trace_id propagated on the envelope but no console UI reads or links on it
• DP-04 alert rules fire to email only — console has no firing-alerts surface
• DP-10 per-bot Prometheus labels not surfaced through BFF

pre-testnet-correctness-gates

Seven stories promoted from backlog in the 2026-05-15 audit. Gates the r-06 testnet trial.

architecture-hardening

14 stories from the NT/RustQuant research wave + pre-cutover P1 gaps.

research-portfolio

Ongoing: 11 strategies under walk-forward validation in gordon-lab.

Cutover sequence

cutover-srv-apps-deploy (Bricks 0-3 live)
    │
    ▼
cutover-staged (Story 24 — live bots)
    │
    ▼
pre-testnet-correctness-gates epic
    │
    ▼
r-06 testnet trial (30 days)
    │
    ▼
op-14 testnet log-review drill
    │
    ▼
production

Key in-flight blockers

• DP-12 console tail: riskWriterClient removal in gordon-console — backend proxy shipped, TS wiring remaining
• gordon-data auto-precompute: bots need higher-TF klines on warmup; WS currently writes 1m only
• node-20 CI deprecation: hard deadline 2026-06-02 across all repos
• observability-console-audit: P0 console observability gaps before testnet