Portfolio Dashboard: ingest real positions from OpenBB into Mira dashboard

[toasterman234]

What I Want

Wire my actual portfolio into Mira's dashboard. My positions live in a local OpenBB Workspace portfolio app at http://localhost:9078. Mira already has the analysis framework (portfolio construction review, position review, thesis cross-referencing) but the private/portfolio/ directory is empty. I want:

• A pipeline that pulls my holdings from the OpenBB endpoint into Mira's private state
• Thesis-to-position cross-referencing (which holdings have Mira cases already?)
• A "Portfolio" view in the Mira Case Cockpit dashboard showing holdings, exposure, concentration warnings, and coverage gaps

Context — The Data

Portfolio widget server at :9078: Schwab (8 positions, ~$29.6K) + Fidelity (20 positions, ~$27.7K) = 28 positions, ~$57.3K total.

Endpoints:

• /schwab/positions, /fidelity/positions — symbol, description, units, price, market_value, open_pnl, avg_cost, day_change, 52w range, sector
• /schwab/summary, /fidelity/summary — total value, P&L, position count
• /schwab/sectors, /fidelity/sectors — dollar/pct allocation by sector
• /schwab/beta, /fidelity/beta — per-position beta + portfolio beta

Mira case overlap: Only 2 of 28 positions have cases:

• EOSE → cases/eose-2026-05/ (Fidelity: $1,287, +$982 P&L, thesis active)
• LB → referenced in cases/royalty-king-four-pillars-2026-07/ (Fidelity: $4,252, +$1,282 P&L)

26 positions have no thesis coverage — these become the natural queue for future Mira research.

Notable concentration: CLM is 46% of the Schwab sleeve ($13,687). EVVTY is 21% of Fidelity ($5,844).

Context — What Mira Already Has

• ✅ Portfolio construction review loop (loops/portfolio-construction-review-loop.md) — exposure map, concentration, duplicate bets, thesis conflicts, stale positions
• ✅ Portfolio review loop (loops/portfolio-review-loop.md) — PM-facing multi-thesis review
• ✅ Position review loop (loops/position-review-loop.md)
• ✅ Portfolio templates (templates/portfolio-system/) — CSV templates for register, exposure review, construction review
• ✅ Dashboard intake actions — "Review a position" and "Review portfolio construction" already in dashboard/choices.json
• ❌ private/portfolio/ — doesn't exist (no holdings, weights, or registers)
• ❌ private/views/view-register.csv — only 1 entry ("AI adopters")
• ❌ No portfolio view in the dashboard (board + cockpit + intake are the 3 current views)

Plan

Phase 1 — Ingest script (scripts/pull_portfolio.py)

• Reads :9078 endpoints into Mira's private state
• Outputs under private/portfolio/:
  • holdings.csv — all 28 positions with Mira-compatible fields (ticker, account, units, price, market_value, open_pnl, avg_cost, sector, date_pulled)
  • summary.json — per-account totals
  • sectors.csv — sector allocation
  • register.csv — Mira portfolio-register template populated
• Idempotent (safe to re-run; overwrites previous snapshot)
• Stdlib-only Python, same convention as existing dashboard/build_data.py

Phase 2 — Cross-reference with Mira cases

• build_data.py extended to read private/portfolio/ and match tickers against existing cases
• Each position gets a thesis_coverage flag: covered / uncovered / partial
• For covered positions, link to the case slug + thesis state + stale status
• Flag concentration risks (any position > 20% of total, any sector > 40%)
• Flag stale theses that overlap with active positions

Phase 3 — Dashboard "Portfolio" view

• New fourth view in the cockpit (alongside Board, Cockpit, Intake)
• Summary bar: total portfolio value, total P&L, split by account, position count
• Holdings table: sortable (by weight, P&L, range %), with thesis-coverage column (green dot = covered, click to open that case's cockpit; gray dot = uncovered, click → intake form)
• Exposure panel: sector breakdown pie/bar (reused from :9078 sector endpoints), beta-weighted concentration map
• Concentration warnings: automatic flag for CLM (46% Schwab), EVVTY (21% Fidelity), and any sector > 40%
• Coverage gap list: the 26 uncovered tickers with a "Start Research" button per position (pre-fills Mira intake with ticker + "from portfolio")
• Data freshness: last pull timestamp, "Refresh from OpenBB" button that calls pull_portfolio.py
• Follow-up queue: positions whose linked thesis is stale → auto-flagged for monitoring loop

What stays unchanged

• Mira's research loops (portfolio data is input, loops don't change)
• The existing case pipeline and cockpit views
• The :9078 portfolio-widget-server (read-only pull, never push)
• No new dependencies — same stdlib+vanilla JS pattern as the existing dashboard