zfin/TODO.md

610 lines
29 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Future Work
No work here is blocking — we're in a good state. Items below are
ordered roughly by priority within each section. Priority labels
(`HIGH` / `MEDIUM` / `LOW`) mark items that deserve explicit
ranking; unlabeled items are "someday, if the mood strikes."
**Next up:** configurable benchmark symbols (low-effort win) and the
manual-check accounts mechanism (medium effort, real user value).
## Projections: future enhancements
- **Configurable benchmark symbols — priority MEDIUM.** Currently
hardcoded SPY + AGG. Route through `projections.srf` as a
`type::config,benchmark::SYMBOL` record (or similar). Low effort.
- **Configurable return cap per position — priority MEDIUM.**
Default: none; cap outliers like NVDA. Should route through
`projections.srf` cleanly.
- **Accumulation-mode SWR rate column is misleading — priority LOW.**
When `retirement_age`/`retirement_at` is configured, the "Safe
Withdrawal" table's % column divides the SWR amount by the
CURRENT portfolio value, not the post-accumulation portfolio
value. The dollar amount is correct (it's the safe spending in
retirement, given the projected accumulation), but the % rate
comes out absurdly high (e.g., 22% of today's portfolio). The
Accumulation phase block already shows the median portfolio at
retirement, so the user can compute the real rate themselves —
but the SWR table's rate column should ideally divide by the
median post-accumulation value, or be suppressed when accumulation
is active. Decide which.
- **Chart vertical line at retirement boundary — priority LOW.**
The accumulation-phase spec called this "mandatory" but it was
explicitly deferred during implementation. The chart currently
shows the full `accumulation_years + horizon` span without a
visual marker for where accumulation ends and distribution
begins. Easier to add to the kitty-graphics chart than the braille
one.
- **Goal-seek over distribution horizon for W1 — priority LOW.**
Today the W1 ("set spending, find date") workflow reports the
earliest retirement at each user-configured `(horizon, confidence)`
cell. The philosophically correct version asks "when have I
accumulated enough wealth that the projection shows a 95%
probability of success withdrawing X per year from retirement
until age-of-death?" — i.e. goal-seek across both `accumulation_years`
AND `distribution_years` simultaneously, anchored to a configured
age-of-death. NP-shaped search; not worth optimizing until
someone wants it.
- **Per-person retirement_age — priority LOW.**
V1 of the accumulation-phase spec chose Option A: a single
household retirement boundary derived from the oldest configured
birthdate. Households where one earner retires significantly
earlier than the other would benefit from per-person
`retirement_age` fields on each `type::birthdate` record, with
contributions stopped per-person.
- **Configurable max_accumulation_years — priority LOW.**
Hardcoded at 50 years. Route through `projections.srf` if anyone
hits the cap.
- Configurable MIN period selection (currently 3Y/5Y/10Y, exclude 1Y)
- Multiple spending models: flat (current), decreasing (1-2% real annual decrease,
Blanchett "spending smile"). Late-life healthcare better modeled as a life event.
- Unclassified position handling in allocation split (warn user)
- **Historical projection overlay follow-ups.** The base
`--overlay-actuals` overlay shipped (CLI tip + TUI primary surface).
Open enhancements:
- Historical `metadata.srf` / `projections.srf` for back-dated
runs. Today the overlay re-runs against current classifications
and assumptions; for historically faithful what-the-model-said-then
output we'd check out the git-tracked versions of those files
at the as-of commit and load those instead. Edge case until
classifications materially drift.
- Contribution-attribution overlay. Today's actuals line includes
contributions implicitly; the bands assume modeled contributions
that may or may not match reality. A "decompose actuals into
market return vs contributions" annotation would clarify how
much of the trajectory was the model being right vs new money
arriving on schedule.
- Mosaic mode: overlay multiple as-of starting points on one chart
("show me 1Y, 3Y, 5Y, 10Y projections all at once") so the user
can see how the projection envelope tightened as data came in.
- **Better composition basis for imported-only as-of.** Today
the imported-only path uses today's allocations scaled by
`imported_liquid / today_total_liquid`. That's the simplest
thing that could work, but it's "today's mix back-dated" —
it ignores everything we know about the historical context.
Specifically: `imported_values.srf` already carries an
`expected_return` field per row that the user captured at
that date in their source spreadsheet. We could:
- Use the imported `expected_return` as a sanity check
against the simulation's per-position weighted return
(warn or clamp if they diverge wildly — the spreadsheet's
number reflects what the user actually saw at the time).
- Use the imported `expected_return` to bias the
stock/bond split inference: a higher expected return
implies a higher historical equity weighting than today's
mix probably reflects.
- Reach further: derive a synthetic stock/bond split from
the imported `expected_return` directly, treating it as
a weighted average of SPY and AGG returns at that date
and solving for the weights. That gives a per-imported-
row composition that's locally faithful instead of
one-mix-fits-all.
None of these are urgent — the current "today's mix scaled"
approximation is documented as such and the bands still
render meaningfully — but each would tighten the historical
faithfulness one notch. Pick whichever has the highest
payoff vs. complexity when this gets revisited.
- **Chart zoom for short-history overlays.** With a 50-year
projection horizon and only ~10 years of imported actuals,
the actuals line is squashed into the first 20% of the
chart and the comparison-against-bands story is hard to
read. Two design directions:
- **Auto-zoom**: when the overlay is on, the chart's
x-axis defaults to `[as_of, today + N years]` (where
N is small, e.g. 2x the actuals span) instead of
`[as_of, as_of + horizon]`. The bands beyond `today
+ N` are still computed but clipped from view. The
tradeoff: the user loses the long-tail terminal-value
context unless they toggle back out.
- **Toggle**: a separate keybind (e.g. `z` for zoom)
flips between full-horizon and zoomed views. Default
off so the bands tell their full story; user opts in
when they want overlay legibility.
Auto-zoom is more invasive (changes the default chart
semantics for everyone running with overlay-on) but better
matches what the user actually wants when they toggle the
overlay. Toggle is safer but requires the user to know the
feature exists. Probably do auto-zoom but expose a toggle
to escape it ("show full horizon").
## Export chart as PNG (`--export-chart <path>`) — priority MEDIUM
z2d already supports PNG export natively. Today the chart-bearing
commands (`quote`, `history`, `projections`, plus the equivalent TUI
tabs) render to braille (CLI) or Kitty graphics (TUI). Adding a
`--export-chart <path>` flag would land just the chart (not the
surrounding text output) as a PNG file at the given path, at full
fidelity, regardless of which surface invoked it.
Driver: when reviewing a back-dated projection or a notable price
move, capturing the chart as an image (e.g. for a write-up, an email
to the household, or a wiki page) is currently a screenshot-and-crop
chore. PNG export makes it a one-shot CLI invocation.
Sketch:
- `zfin quote AAPL --export-chart aapl.png` → just the price+
Bollinger chart as a PNG, no other output.
- `zfin projections --as-of 1Y --overlay-actuals --export-chart projection.png`
→ the projection-bands chart plus overlay, no other output.
- The chart code already produces RGB pixel buffers via z2d; replace
the `transmitPreEncodedImage` call (TUI) or the braille text path
(CLI) with a `Surface.write_png` call when the flag is present.
Plumbing: a thin "chart-only render" entry point in each chart
module (`projection_chart.zig`, `chart.zig` for symbols), called
from the relevant command's `run()` when `--export-chart` is set.
Exits before the rest of the text output renders.
Out of scope for V1: file-format alternatives (SVG, PDF), themed
color overrides for export (always uses the active terminal theme),
non-chart command output as PNG.
## Refactor: trim `src/format.zig` once Money / Date have absorbed their helpers — priority LOW
`src/format.zig` is still a ~1700-line grab-bag, but the money- and
date-shaped helpers that used to live there have been moved out:
money formatting now lives in `src/Money.zig` (with `{f}` /
`whole()` / `trim()` / `signed()` / `padRight(N)` / `padLeft(N)`),
and date formatting lives in `src/Date.zig` (with `{f}` /
`padRight(N)` / `padLeft(N)`). What's left in `format.zig` is the
genuinely-format-domain stuff: braille charts, return formatters,
allocation notes, signed-percent rendering.
If the file ever grows enough to be annoying again, consider
renaming to `src/render.zig` to better describe what's left, or
splitting the braille chart out (it's ~600 lines on its own).
Not blocking — file it as cleanup if and when it bites.
## Refactor: TUI App struct knows too much about each tab — priority MEDIUM
`src/tui.zig`'s `App` struct currently has dozens of tab-specific
fields scattered across its top level — `projections_loaded`,
`projections_disabled`, `projections_config`, `projections_ctx`,
`projections_horizon_idx`, `projections_image_id`,
`projections_image_width`, `projections_image_height`,
`projections_chart_dirty`, `projections_chart_visible`,
`projections_events_enabled`, `projections_value_min`,
`projections_value_max`, `projections_as_of`,
`projections_as_of_requested`, `projections_overlay_actuals`,
plus equivalents for portfolio, history, options, earnings,
analysis, perf, and quote tabs.
This couples `App` to the implementation details of every tab.
Touching a single tab's state shape requires editing the central
struct, which makes refactors noisy and discourages tab modules
from owning their own state cleanly.
**Proposal: each tab gets exactly ONE field on `App`** — a
struct (or pointer to a struct) defined in that tab's own file.
The tab module owns its state shape; `App` only carries the
top-level reference.
Sketch:
```zig
// src/tui/projections_tab.zig
pub const State = struct {
loaded: bool = false,
disabled: bool = false,
config: projections.UserConfig = .{},
ctx: ?ProjectionContext = null,
horizon_idx: usize = 0,
chart: ChartState = .{},
as_of: AsOfState = .{},
overlay_actuals: bool = false,
// ...
};
// src/tui.zig
pub const App = struct {
// ... shared cross-tab state (today, allocator, vx_app, ...) ...
portfolio: portfolio_tab.State = .{},
quote: quote_tab.State = .{},
perf: perf_tab.State = .{},
history: history_tab.State = .{},
projections: projections_tab.State = .{},
options: options_tab.State = .{},
earnings: earnings_tab.State = .{},
analysis: analysis_tab.State = .{},
};
```
After the migration, `app.projections_overlay_actuals` becomes
`app.projections.overlay_actuals`, etc. Each tab's State struct
documents its own invariants without polluting `App`.
**Benefits:**
- Tab modules become genuinely self-contained. Adding new state
to a tab only touches that tab's file.
- `App`'s field count drops from ~80 to ~10 (cross-cutting state
+ 8 tab-state fields).
- Clearer ownership: when reading `App`, you can tell at a
glance which fields are shared vs tab-private.
- Easier to reason about lifetimes — each tab's `freeLoaded()`
operates on its own struct rather than reaching into App.
- Onboarding new contributors: "to add a tab feature, define
state in the tab file" instead of "edit two files in lockstep."
**Migration approach:**
This is a large mechanical refactor (~80+ field renames across
all tab files plus `tui.zig`). Best done as one focused PR:
1. Define `State` in each tab file with all current fields.
2. Add the eight new fields to `App`; flip them on as defaults
(so old `app.projections_*` accesses temporarily still work
if we add accessor shims, but cleaner to just rip the
bandaid).
3. Sweep all `app.<tab>_<field>` references → `app.<tab>.<field>`.
4. Delete the old top-level fields from `App`.
5. Verify `zig build test` is unchanged. The refactor should be
strictly behavior-preserving.
Risks: large diff (touches every tab file plus tui.zig), but
mechanical — no logic changes, no tests should move. The pre-
commit hooks catch any miss instantly.
**While we're at it: action handler bodies should also move.**
The same shape problem shows up in `tui.zig`'s keybind-action
dispatch — `sort_reverse`, `toggle_chart`, `toggle_events`,
`account_filter`, and the per-tab branches inside them are
~100 lines of tab-specific logic living in the central event
loop. Concretely (line numbers as of writing):
- `sort_reverse` (~line 1328) dual-dispatches by active tab:
portfolio flips sort direction + calls
`sortPortfolioAllocations` / `rebuildPortfolioRows`;
projections toggles overlay-actuals + reloads data + sets
status. None of that body is `App`-level concern; it's two
separate tabs' private state mutations.
- `toggle_chart` (~line 1382) flips
`projections_chart_visible` and resets `scroll_offset`. Pure
projections-tab business.
- `toggle_events` (~line 1390) flips
`projections_events_enabled` and triggers a reload. Same.
- `account_filter` (~line 1366) opens the account picker mode
with portfolio-tab-specific cursor positioning.
The cleaner shape: each tab module exposes a
`handleAction(app, action) bool` (or similar) that returns
true when it consumed the action. `tui.zig`'s dispatch becomes
a thin "ask the active tab if it wants this action; otherwise
fall through to global handlers." The body of each `case`
shrinks to a one-liner that dispatches to
`tab_module.handleAction(self)`.
This pairs naturally with the State-struct migration above —
the tab module's `handleAction` operates on its own State
struct rather than reaching into `App`. Some keybinds are
genuinely cross-cutting (quit, refresh, tab navigation, scroll)
and stay in the central handler. The split is "App owns
chrome; tab owns content."
Driver: every overlay-actuals-shaped feature added to a tab
recently has involved adding 12 fields to `App`, and the
struct keeps growing. Eventually it becomes unreviewable.
## Bug: braille charts use raw `close`, not `adj_close` — cliff at splits
**Reproduction:** `zfin quote SOXX` (or the TUI quote tab). The
braille chart drops sharply on **2024-03-07**, which is the
iShares Semiconductor ETF's 3-for-1 split date:
- 2024-03-06 close: $689.60
- 2024-03-07 close: $237.75 (≈ $689.60 / 2.9)
The `adj_close` column in `~/.cache/zfin/SOXX/candles_daily.srf`
tracks correctly through the split (~$226 → ~$234), so the
provider data is fine. The bug is purely cosmetic: the chart
renders the *unadjusted* close price.
**Root cause:** `computeBrailleChart` in `src/format.zig:888`
indexes `data[i].close` instead of `data[i].adj_close`. Lines 901,
902, 904, 905, 935 all use `.close`.
**Independent confirmation:** `zfin splits SOXX` returns
`2024-03-07 3:1` from Polygon. So the split data exists in the
provider layer (and gets cached as `splits.srf` once requested),
but the charting code path doesn't consult it.
**Fix candidates:**
A. **Switch `computeBrailleChart` to consume `adj_close` directly.**
Simplest. Affects every chart caller (quote, history, projections
median band, TUI quote/projections tabs). Cosmetic only — no
computation depends on it. The price-axis labels would render
adjusted prices, which may surprise users used to seeing the
raw last-close. Mitigate with a comment in the chart's right-edge
label region or a header note.
B. **Pass a flag to `computeBrailleChart` selecting `close` vs
`adj_close`.** Default to adjusted; let the quote tab show raw.
More flexible, marginally more code.
C. **Add a `chart_close` accessor to `Candle` that returns
`adj_close` if non-zero, else `close`.** Same effect as (A) with a
defensive fallback.
D. **Apply split adjustments at chart-data prep time using
`splits.srf`.** Walk the candle slice with the split history and
pre-multiply pre-split closes by the cumulative ratio. More
work, but produces a chart-axis dollar value the user expects
("today's last close was $X, the chart starts at $Y from N
years ago"). This is what most charting libraries do.
Requires plumbing `DataService.getSplits` into the chart-prep
path on every chart caller, OR doing the adjustment once in the
service layer alongside candle fetching. Not all callers have a
`DataService` reference today (e.g., `runProjection`'s synthetic
median-band candles).
**Recommendation:** Start with (A) or (C) — single-line fix, gets
the cliff out of all charts immediately. (D) is the "correct" fix
but a bigger refactor; file as a follow-up if (A)/(C) lands first.
**Other affected symbols:** Any held position with a split in the
last 10 years will have the same artifact. Check NVDA (10:1 split
on 2024-06-10) for a louder example.
**Priority:** LOW. Cosmetic only — analytics already use
`adj_close` correctly via the per-position trailing-returns path.
But it's confusing when scanning a chart and seeing a 50% drop
that isn't real.
## Audit: manual-check accounts mechanism (NYL, Kelly's ESPP, etc.) — priority HIGH
Some accounts/positions can't be reconciled from broker CSVs and need a
human-in-the-loop reminder at the audit step. Examples:
- **NY Life** — no CSV export at all. Values only live in periodic
statements.
- **Kelly's ESPP** — accrued payroll-deduction cash doesn't appear in the
Fidelity positions CSV until the purchase date hits (typically every
6 months). Between purchases the cash is a real contribution that
`zfin audit` can't see.
- Future: treat as an open category.
The existing `update_cadence::weekly|monthly|quarterly|none` field already
sort-of covers this, but has two gaps:
1. It fires off the last *git-detected change*, not the last *human
review*. For NYL, the value sometimes hasn't changed in months — so
git never fires, cadence never trips.
2. ESPP needs weekly-ish attention while accumulating cash between
purchases, but the accrued balance is invisible to the CSV audit.
### Options
A. **New `update_cadence::manual` variant** — always fires every audit
run until silenced. Blunt but zero design work.
B. **`last_refreshed::YYYY-MM-DD` field on `accounts.srf`** — explicit
human-review timestamp, decoupled from git-detected changes. Audit
compares `today - last_refreshed` against the cadence. User bumps
the field when they check the statement. Probably the most
correct fit for NYL.
C. **Sticky TODO list** — a `todos.srf` or `todo::` field on accounts
that audit always surfaces until cleared. General-purpose; also
covers "remember to rebalance on 5/15".
### ESPP-specific follow-through
ESPP is also a contribution-attribution blind spot. If Kelly's paystub
deducts $X/week but the cash lot doesn't reach `portfolio.srf` until
the purchase date, the attribution math is under-counting contributions
and over-counting the purchase-week gain. Possible fixes are discussed
in the "Contributions diff" TODO below — option C there (per-account
`cash_is_contribution`) would make manually-entered ESPP
cash additions count correctly.
## In-kind transfer support (`type::in_kind`) — priority MEDIUM
`transaction_log.srf` parses `type::in_kind` records but the
contributions matcher always rejects them with "in-kind transfers
not yet supported in v1." In-kind movements need per-symbol
matching across accounts: an in-kind transfer of 100 VTI shares
from Acct A to Acct B shows up as `lot_removed` on A + `new_stock`
on B (or a `rollup_delta` share increase if B already had a VTI
lot), neither of which can be matched by the current
amount-based cash matcher.
Proposed: a second pass in `matchTransfers` that iterates
`type::in_kind` records and looks for same-symbol matches across
`lot_removed` on `from` + `new_stock`/`rollup_delta` on `to`
within the window. Gated on share-count and open_price sanity so
a partial transfer doesn't false-positive against an unrelated
edit.
Driver: when the user starts moving positions between accounts
directly (e.g. Roth conversion of already-held shares, 401k →
rollover IRA in-kind) rather than liquidating and re-buying.
## Torn SRF files from server sync (root cause unknown)
**Status:** Root cause still unidentified. We have mitigations and
diagnostics in place that keep torn responses from corrupting the
cache, but we don't yet know *why* responses arrive torn. Until we
have a root cause, this is not resolved — it's mitigated.
Mitigations landed so far:
- `syncFromServer` (`src/service.zig`) validates responses via
`cache.Store.looksCompleteSrf` before `writeRaw`. Torn HTTP bodies
(empty, missing `#!srfv1` header, or no trailing newline) are
rejected with a warn-level log and NOT written to cache.
- HTTP responses are checked for an `ETag` sha256 header; on mismatch
we retry the request once before giving up and falling back to the
provider.
- Read-path self-heal: on SRF parse failure during read, the cache
entry is invalidated so a subsequent refresh can repair without
user intervention.
- Diagnostics: richer error capture around the sync path. So far,
HTTP transit is the dominant source of torn responses — but that's
an observation, not a root cause.
**Remaining work:**
- Identify root cause. Candidates to investigate: proxy/load-balancer
behavior, HTTP keepalive reuse, partial reads on the server side,
client-side buffer handling. The etag retry tells us whether the
problem is per-request or persistent; dig into the diagnostics
output when the next occurrence is captured.
- Once root cause is known, decide whether the current mitigations
are sufficient or whether a targeted fix is needed. The
mitigations may end up being the whole answer, but we can't
conclude that without understanding the underlying cause.
(Content-Length validation was considered and rejected: once the
server starts compressing response bodies, Content-Length reflects
the compressed byte count, not the decoded payload, so it's not a
reliable integrity check.)
## Market-aware cache TTL for daily candles
Daily candle TTL is currently 23h45m, but candle data only becomes meaningful
after the market close. Investigate keying the cache freshness to ~4:30 PM
Eastern rather than a rolling window. This would avoid unnecessary refetches
during the trading day and ensure a fetch shortly after close gets fresh data.
Probably alleviated by the cron job approach.
## On-demand server-side fetch for new symbols
Currently the server's SRF endpoints (`/candles`, `/dividends`, etc.) are pure
cache reads — they 404 if the data isn't already on disk. New symbols only get
populated when added to the portfolio and picked up by the next cron refresh.
Consider: on a cache miss, instead of blocking the HTTP response with a
multi-second provider fetch, kick off an async background fetch (or just
auto-add the symbol to the portfolio) and return 404 as usual. The next
request — or the next cron run — would then have the data. This gives
"instant-ish gratification" for new symbols without the downsides of
synchronous fetch-on-miss (latency, rate limit contention, unbounded cache
growth from arbitrary tickers).
Note that this process doesn't do anything to eliminate all the API keys
that are necessary for a fully functioning system. A more aggressive view
would be to treat ZFIN_SERVER as a 100% source of record, but that would
introduce some opacity to the process as we wait for candles (for example) to
populate. This could be solved on the server by spawning a thread to fetch the
data, then returning 202 Accepted, which could then be polled client side. Maybe
this is a better long term approach?
## Low-priority items
The following items are acknowledged but not prioritized. Listed here
so they don't get lost; pick up opportunistically.
### UX
- **CLI options command UX.** The `options` command auto-expands only
the nearest monthly expiration and lists others collapsed. Reconsider
the interaction model — e.g. allow specifying an expiration date,
showing all monthlies expanded by default, or filtering by strategy
(covered calls, spreads).
- **TUI: toggle to last symbol keybind.** A single-key toggle that
flips between the current symbol and the previously selected one
(like `cd -` in bash or `Ctrl+^` in vim). Store `last_symbol` on
`App`; on symbol change, stash the previous. Useful for
eyeball-comparing performance/risk data between two symbols.
### Data quality
- **Fix `enrich` command for international funds.** `deriveMetadata`
in `src/commands/enrich.zig` misclassifies international ETFs:
1. `geo` uses Alpha Vantage's `Country` field, which is the *fund
issuer's* domicile (USA for all US-listed ETFs), not the fund's
investment geography. Every US-domiciled international fund gets
`geo::US`.
2. `asset_class` short-circuits to `"ETF"` when
`asset_type == "ETF"`, or falls through to a US-market-cap
heuristic that always produces `"US Large Cap"` /
`"US Mid Cap"` / `"US Small Cap"`.
Known misclassified tickers (all came back as
`geo::US, asset_class::US Large Cap`):
- **FRDM** — Freedom 100 Emerging Markets ETF → should be
`geo::Emerging Markets, asset_class::Emerging Markets`
- **HFXI** — NYLI FTSE International Equity Currency Neutral ETF
→ should be
`geo::International Developed, asset_class::International Developed`
- **IDMO** — Invesco S&P International Developed Momentum ETF →
should be
`geo::International Developed, asset_class::International Developed`
- **IVLU** — iShares MSCI International Developed Value Factor
ETF → should be
`geo::International Developed, asset_class::International Developed`
The Alpha Vantage OVERVIEW endpoint doesn't provide fund geography
data. Options: use the ETF_PROFILE holdings/country data to infer
geography, parse the fund name for keywords ("International",
"Emerging", "ex-US"), or accept that `enrich` is a scaffold and
emit a `# TODO` comment for ETFs instead of silently misclassifying.
### Options / valuation
- **Per-account covered call adjustment.** `adjustForCoveredCalls` in
`valuation.zig` operates on portfolio-wide aggregated allocations.
It matches sold calls against total underlying shares across all
accounts. This is wrong — calls in one account can only cover
shares in that same account. Fixing means restructuring
`portfolioSummary`, since `Allocation` is currently
account-agnostic. Low priority — naked calls are rare, and calls
are typically in the same account as the underlying.
- **Covered call adjustment O(N*M) loop.** `adjustForCoveredCalls`
has a nested loop — for each allocation, it iterates all lots to
find matching option contracts. Fine for personal portfolios
(<1000 lots). Pre-indexing options by underlying would help if
someone had a very large options-heavy portfolio.
### Analysis / correctness
- **Analysis account/asset-class total mismatch.** The "By Account"
and "By Tax Type" sections in the analysis command sum to slightly
more than "Asset Class" (~0.6% error). Likely a discrepancy between
how the lot-level account loop values cash, CDs, or options vs how
the asset-class section computes them via `portfolio.totalCash()` /
`totalCdFaceValue()`. Per-account values themselves are correct
after the price_ratio fix.
### Audit
- **Audit large-lot threshold tuning.** `src/commands/audit.zig` uses
`audit_large_lot_threshold: f64 = 10_000.0` as the cutoff for
"surface this new lot for confirmation." Revisit if $10k proves too
aggressive (ESPP accruals spam the report) or too permissive (large
DRIP confirmations slip past). If runtime tuning becomes necessary,
a `--large-lot <amount>` flag or a global
`audit_large_lot_threshold` field on `accounts.srf` would be
reasonable extensions.
### Infra / performance
- **HTTP connection pooling.** Parallel server sync in `loadAllPrices`
spawns up to 8 threads, each with its own HTTP connection. Could
reuse connections to reduce TCP handshake overhead. Only matters
with very large portfolios (100+ symbols) hitting ZFIN_SERVER.
- **Streaming cache deserialization.** Cache store reads entire files
into memory (`readFileAlloc` with 50MB limit). For portfolios with
10+ years of daily candles, this could use significant memory.
Keep current approach unless memory becomes a real problem.