Epic: Wheeler phase 2 — multi-tenant + IBKR Flex

[heyitsStylez]

Tracks Wheeler's phase-2 work, deferred and hard-gated in the v1 spec (#86). Lighter-weight than the v1 wayfinder map (#78) — the shape is already largely decided; this is one decision plus its build fallout.

Two deferred tracks (from #86)

• Multi-tenant auth + per-user server storage (Decision: account/identity model replacing wallet-based cloud sync #84) — the gate.
• IBKR Flex auto-import (Research: IBKR data ingestion feasibility (Client Portal API vs Flex Query) #79/Decision: manual-input v1 vs IBKR API pull, and phasing #83) — rides behind the gate; needs per-user Flex tokens, which need auth + per-user storage.

Critical path

Everything hinges on the identity/storage decision — Flex tokens are per-user secrets that can't exist until auth + a per-user datastore do.

• Decision: identity + storage backend — Decision: identity + storage backend for Wheeler phase 2 #105 (auth model → datastore → Flex-token storage, resolved as one cluster)

Build tickets below get created once #105 lands (their shape depends on it):

• Auth + per-user persistence impl (Wheeler phase 2: auth + per-user persistence impl (Google OAuth session) #110) — swap Wheeler's local tradfi/12a-persistence.js seam for the authenticated server seam; currentUserKey() returns the real user id. core/ untouched.
• api/ibkr-flex.js (Wheeler phase 2: api/ibkr-flex.js — Activity Flex import + encrypted per-user token #111) — Activity Flex proxy (research Research: IBKR data ingestion feasibility (Client Portal API vs Flex Query) #79: Vercel serverless, token-based, T+1 EOD accepted). Dedup/import mirroring the existing chain-apply pattern (18b-chain-apply.js).
• Automated early-assignment detection (Wheeler phase 2: automated early-assignment detection from Flex #112) — rides in with Flex per Decision: manual-input v1 vs IBKR API pull, and phasing #83 (manual ASSIGNED-with-date is the v1 story).

Settled going in

• No interim single-user Flex (Decision: manual-input v1 vs IBKR API pull, and phasing #83).
• Persistence seam already async and ready (Decision: adapt hyperwheel in-place vs new sibling repo #82/Decision: account/identity model replacing wallet-based cloud sync #84) — server backend slots in without a core/ refactor.
• Flex ingestion path researched (Research: IBKR data ingestion feasibility (Client Portal API vs Flex Query) #79); equity price feed already shipped in v1 (Research: equity/ETF live price feed for spot marks #80).

Known v1 gaps (may fold in here or stay separate)

• Partial PUT assignment (Domain delta: lot engine under ×100 multiplier & share-based assignment #81).
• Dynamic per-ticker filter tabs (Wheeler: dynamic per-ticker filter tabs (arbitrary tickers) #97), build-config duplication (Wheeler build: fragment/app config duplicated between build.py and test/helpers/assemble.js #98) — v1 tail, not phase 2.

[heyitsStylez]

v1 tail (tracked here for visibility, not part of phase 2): #97 dynamic per-ticker filter tabs, #98 build.py/test fragment-config duplication.