Appearance
Implementation Roadmap
This page tracks what is shipped, what diverges from the target product shape, and what should come next. Detailed API and table contracts live in the linked docs.
Related:
- Frontend Integration Guide
- Backend Overview
- Clash Of Clones Product Spec
- Backend Contract
- Backtesting Module
- Prop Trading Module
- Lifecycle Module
- Product Controls
Backend Ownership Split
The current backend is organized around modules, not planning silos:
| Module | Owns | Does not own |
|---|---|---|
| Data Ingestion | Global market memory, source health, asset registry, derived indicators | Per-clone strategy or trading decisions |
| AI Trading | Strategy compilation, context packets, prompt/model execution, proposed actions | Balances, fills, lifecycle outcomes, payment facts |
| Backtesting | Historical/scenario replay runs, simulated clocks, daily usage limits, run artifacts, comparative metrics | Live prop-account truth, lifecycle settlement, payment facts |
| Prop Trading | Prop accounts, safety gates, paper fills, positions, PnL, ledger truth | Race settlement, tickets, creator payouts |
| Economy V1 (Audition) | Audition-fee payment verification, purchase facts, entitlement grants, voided-audition withdrawal signatures | Clone strategy, race ranking, competition settlement |
| Lifecycle V1 (Audition) | Paid audition submit, evaluation, retry, void, and run status | Execution mechanics, prop-account ledger truth, or competition lifecycle |
| Competition Manager V1 (Arena Entry) | Promotion Queue, Arena slot admission, and promotion from passing auditions | Prop-account execution truth, Economy-owned payment/payout facts, race cadence, tickets, settlement, retirement, or Hall of Fame |
Lifecycle V1 and future Competition Manager work should create and consume prop accounts through explicit module APIs. They should not reach into prop trading tables to mutate fills, positions, balances, or ledger rows.
Backtesting should reuse AI Trading decision execution and Prop Trading fill math, but it should write to backtest-owned jobs, artifacts, and metrics instead of production prop-account tables.
Backend Progress By Module
Percentages are rough v1 backend completion estimates, not release gates.
| Module | Complete | Done | Pending |
|---|---|---|---|
| User/Auth | ✅ 100% | SIWE challenge/verify, JWT sessions, refresh rotation, dev fallback, user/wallet resolution. | Prod prep only. |
| Data Ingestion | ✅ 100% | Hyperliquid, Polymarket, DefiLlama storage; scheduled feature channels; source health APIs; local retry/backfill tooling. | Prod prep only. |
| Clone Service | ✅ 100% | Clone create/list/read/update, drafts/versions, graph persistence, layout metadata, entitlements, owner public profile, public reporting, admin moderation. | Frontend integration only. |
| AI Trading | ✅ 100% | Context resolver, prompt v1, provider routing, decision rows/actions, workflow dry run, prop-account per-asset cadence/audit, prop-account action submission, artifact storage, and topology entitlement guardrails across save, dry-run, backtest, and runtime paths. | Prod prep only. |
| Backtesting | ✅ 100% | Dedicated module shell, durable async job state, worker queue leases, global/per-user concurrency controls, comparable metrics, usage-day/artifact tables, artifact retention, Prop Trading fill-math reuse, tier-aware daily quota, explicit preflight/history APIs, and per-run cap checks; clone-scoped route remains API-compatible. | Frontend history/comparison UI and final pricing policy. |
| Prop Trading | ✅ 100% | Prop accounts, safety gates, paper fills, balances, positions, snapshots, mark-to-market, admin controls. | Moved to Lifecycle. |
| Economy V1 (Audition) | ✅ 100% | Audition-fee quote/submit, SignedVault verification, purchase idempotency, entitlement grant. | Prod prep only. |
| Lifecycle V1 (Audition) | ✅ 100% | Submit/evaluate/retry/void on paid auditions; shared audition contract boundary; owner-safe routes; due-audition evaluator worker. | Backend-complete for Audition V1; frontend status/retry UI and production evaluator monitoring are tracked separately. |
| Competition Manager V1 (Arena Entry) | ✅ 100% | Promotion Queue and Arena membership tables, public Arena/Queue read routes, service-token sync, worker cadence, approved-profile admission gates, and admin removal/retirement controls. | Backend-complete for Arena Entry V1; race launch belongs to later competition lifecycle work. |
| Operations/Workers | 78% | Worker entrypoints, migrations, heartbeat storage, /health/workers, ingestion health APIs, backtest queue leases, and Arena Entry cadence. | Retries, dead-letter handling, alerts. |
Frontend Progress By Surface
Percentages are rough frontend completion estimates against the current product/backend direction.
| Surface | Complete | Done | Pending |
|---|---|---|---|
| App shell / navigation | 60% | Route groups, protected/public layouts, landing/home surfaces, with-nav shell. | Pump Party session gating. |
| Auth UX | 30% | Privy provider and login UI exist. | SIWE session flow. |
| Clone workflow builder | 65% | Workflow canvases, clone strategy graph UI, data-source controls, dry-run/backtest-oriented calls. | Create/list UI against the Clone Service APIs, polish, auth wiring. |
| Backtesting UX | 5% | Existing workflow backtest calls can start capped evaluation jobs against the completed backend. | User-facing job history, run results, metrics comparison, and capped configuration controls. Keep call/token estimates in debug/admin views. |
| Trading dashboard | 45% | Trade page, clone controls, logs, state polling, chart/store scaffolding. | Prop snapshots, runtime controls. |
| Audition + economy flow | 10% | Shared contracts and backend endpoints are documented. | Payment, submit, status, retry UI. |
| Public profile + moderation | 45% | Public profile/profile-action surfaces and social actions exist in legacy shape. | Wire owner profile submit/list/read and report flows to backend. |
| Arena / races / tickets | 10% | Arena/Queue backend reads exist; product docs and shared lifecycle contract shape exist. | Arena/Queue display; race, ticket, and settlement UI after race backend work begins. |
| Ops/admin surfaces | 35% | Ingestion admin/status surfaces exist. | Worker health, reconciliation, audit views. |
Production Readiness
Percentages are rough prod-prep completion estimates. They are separate from backend module correctness.
| Area | Complete | Done | Pending |
|---|---|---|---|
| Runtime process model | 55% | API and worker entrypoints exist; runbook lists process split. | Supervisor/deploy wiring. |
| Secrets / config | 45% | Env contracts documented; dev fallback blocked in production. | Managed secrets, prod env audit. |
| Monitoring / alerting | 35% | Ingestion health, worker heartbeats, health routes. | Host alerts, alert routing. |
| Worker resilience | 40% | Long-running workers, one-shot commands, and backtest leases exist. | Retries, dead-letter handling. |
| Backups / restore | 15% | SQLite/R2 direction documented. | Backup jobs, restore drill. |
| Reconciliation | 20% | Purchase idempotency and audit rows exist. | Audition-purchase, payout, and ledger checks. |
| Admin / incident tools | 45% | Prop admin controls, moderation routes, audit logs. | Operator UI, incident runbooks. |
| Cost / model ops | 35% | Token usage and model routing exist. | AI decision pricing, backtesting pricing, budgets, alerts, provider policy. |
Known Divergence
| Divergence | Why it exists | Direction |
|---|---|---|
| Clone-scoped dry-run execution tables still exist | They support workflow debug and older decision history paths. | Keep for debug; new product runtime should target prop accounts. |
Backtest routes remain clone-scoped under decision-backtest paths | The first evaluation path grew out of workflow dry-run behavior and the frontend already calls this route. | Keep API compatibility while Backtesting owns jobs, usage metering, cost controls, and artifacts behind the route. |
| AI and backtesting pricing is not finalized | AI trading limits are topology-based and now enforced; backtesting cost still depends on replay steps, decision calls, token usage, artifact retention, and user entitlement. | Keep AI topology values and current Backtesting tier limits as configurable guardrail defaults until Economy/Entitlements owns final pricing and overage policy. |
| Lifecycle shared contract includes race/ticket endpoints that are not implemented | Frontend needs stable type and route boundaries before the next product slice. | Treat these as contract-first surfaces; implement behind the existing route names. |
| Mock Starter entitlement fallback still exists | Local/dev and legacy clone paths need a non-payment fallback. | Paid audition flows should rely on Economy-granted entitlements. |
| Public profile moderation is report-based, not pre-approval | Submit auto-publishes; users report; admins retire or reinstate. | Frontend copy should not promise pre-audition moderation review. |
| Race economy exists in docs/contracts before code | Pool math, ticket positions, payouts, and royalties need a stable target. | Implement with Economy owning payment facts and Lifecycle owning race settlement rows. |
Implementation Trajectory
- Frontend integration for Pump Party SIWE sessions, audition-fee quote/submit, paid audition submit, audition status, retry, and public profile/report flows.
- Backtesting frontend controls: job history, run results, metrics comparison, and capped configuration UI.
- Arena/Queue frontend display and operator views for the completed Arena Entry backend.
- Race lifecycle tables and workers: accumulating, launched, live, settling, settled, archived.
- Race-mode prop account creation and result snapshots for every clone in a race.
- Economy V2 ticket quote/submit, race ticket positions, and production reconciliation.
- Settlement: 5% protocol rake, 5% creator pool, 90% ticket-holder distribution, weighted parimutuel payouts, payout authorization rows.
- Retirement strikes, career stats, and Hall of Fame preservation.
- Production readiness: alerts, backup verification, reconciliation dashboards, incident runbooks, cost/token monitoring.
Backtesting Cost Controls
Backtesting must be constrained before it is exposed as a normal user workflow because historical replay can multiply provider calls and token cost.
| Control | Current state |
|---|---|
| Daily quota | Implemented as accepted runs per userId + cloneId + UTC day, with limits derived from the clone's effective entitlement tier. |
| Per-run estimate | Implemented before inline or async start for replay steps, decision ticks, active assets, decision calls, and token budget; exposed through the explicit preflight API for debug/admin/operator views. |
| Replay step cap | Implemented for the current last_6h/5-minute replay shape. |
| Asset cap | Implemented from effective entitlement tier and per-run policy. |
| Model tier policy | Partially covered by clone entitlement/model-strength checks; pricing-specific model policy remains open. |
| Queue and concurrency limits | Implemented through the backtest worker with leased jobs plus global and per-user concurrency limits. |
| Artifact retention | Implemented for backtest artifacts. Summaries and metrics are durable by default; prompt/context/response artifacts expire on the retention window. |
Current Backtesting guardrail defaults:
| Tier | Daily accepted runs | Max decision calls / run | Max active assets / run | Max estimated tokens / run |
|---|---|---|---|---|
| Starter | 3 | 24 | 3 | 120,000 |
| Builder | 10 | 100 | 6 | 500,000 |
| Pro | 25 | 200 | 10 | 1,000,000 |
Out Of Scope For This Roadmap
- Live exchange order placement.
- User personal-wallet balances.
- WebSocket/SSE delivery for frontend polling surfaces.
- Multiple payment assets beyond ETH.
- A second insight-generation layer between market memory and prompts.