Slim more agent output, fix spending report dates, add an output size benchmark

.gitignore

@@ -3,3 +3,4 @@ dist
 coverage
 *.log
 CLAUDE.local.md
+.bench/

CLAUDE.md

@@ -25,6 +25,14 @@ WF_TEST_REFRESH=1 pnpm test:integration
 scripts/test-server.sh down
 ```
 
+Output size benchmark (see README "Output size benchmark"). Run it before and after a change to what a command prints:
+
+```sh
+pnpm build && eval "$(scripts/test-server.sh up)"
+scripts/bench-seed.sh           # once per server; WF_TEST_PROXY=1 on test-server.sh if quotes need a proxy
+scripts/bench.sh main           # Markdown table: main vs dist/wf.js
+```
+
 CI also fails if `vendor/`, `src/contract.ts` or `UPSTREAM_TAG` differ from what is committed (`git diff --exit-code`).
 
 ## Architecture

scripts/bench-seed.sh

@@ -0,0 +1,125 @@
+#!/usr/bin/env bash
+# Seeds a fresh test server (scripts/test-server.sh up) with a fixed data set
+# for scripts/bench.sh:
+#
+#   Brokerage  USD securities: 6 stocks bought monthly in 2025, dividends
+#   Euro ISA   EUR securities: 2 stocks bought quarterly
+#   Checking   USD cash, tracked for spending: November 2025 salary,
+#              categorised spending, and budget targets for 2025-11
+#
+#   eval "$(WF_TEST_PROXY=1 scripts/test-server.sh up)"
+#   pnpm build && scripts/bench-seed.sh
+#
+# Run it once per server: it adds the same data again on a second run. Prices
+# need the server to reach Yahoo Finance; without them, holdings carry nulls
+# and the numbers are smaller than a real portfolio's.
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+: "${WF_TEST_URL:?Run: eval \"\$(scripts/test-server.sh up)\"}"
+: "${WF_TEST_PASSWORD:?Run: eval \"\$(scripts/test-server.sh up)\"}"
+export WF_CONFIG_DIR="${WF_BENCH_CONFIG_DIR:-$ROOT/.bench}"
+mkdir -p "$WF_CONFIG_DIR"
+wf() { node "$ROOT/dist/wf.js" "$@"; }
+id() { node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>console.log(JSON.parse(s).id))'; }
+
+WF_PASSWORD="$WF_TEST_PASSWORD" wf login --url "$WF_TEST_URL" >/dev/null
+wf call update_settings --args \
+  '{"settingsUpdate":{"baseCurrency":"USD","timezone":"UTC","onboardingCompleted":true}}' >/dev/null
+
+account() {
+  wf call create_account --args "{\"account\":{\"name\":\"$1\",\"accountType\":\"$2\",\"currency\":\"$3\",
+    \"isDefault\":false,\"isActive\":true,\"isArchived\":false,\"trackingMode\":\"TRANSACTIONS\"}}" | id
+}
+activity() { wf call create_activity --args "{\"activity\":$1}" | id; }
+
+BROKER=$(account Brokerage SECURITIES USD)
+ISA=$(account "Euro ISA" SECURITIES EUR)
+CHECKING=$(account Checking CASH USD)
+echo "accounts: $BROKER $ISA $CHECKING" >&2
+
+activity "{\"accountId\":\"$BROKER\",\"activityType\":\"DEPOSIT\",\"activityDate\":\"2025-01-02\",\"amount\":\"200000\",\"currency\":\"USD\"}" >/dev/null
+activity "{\"accountId\":\"$ISA\",\"activityType\":\"DEPOSIT\",\"activityDate\":\"2025-01-02\",\"amount\":\"100000\",\"currency\":\"EUR\"}" >/dev/null
+for symbol in AAPL MSFT NVDA GOOGL AMZN VTI; do
+  for month in 01 02 03 04 05 06 07 08 09 10 11 12; do
+    activity "{\"accountId\":\"$BROKER\",\"activityType\":\"BUY\",\"activityDate\":\"2025-$month-15\",
+      \"quantity\":\"2\",\"unitPrice\":\"$((100 + 10#$month * 3))\",\"fee\":\"1\",\"currency\":\"USD\",
+      \"asset\":{\"symbol\":\"$symbol\",\"quoteCcy\":\"USD\"}}" >/dev/null
+  done
+done
+for symbol in SAP.DE ASML.AS; do
+  for month in 02 05 08 11; do
+    activity "{\"accountId\":\"$ISA\",\"activityType\":\"BUY\",\"activityDate\":\"2025-$month-10\",
+      \"quantity\":\"5\",\"unitPrice\":\"200\",\"fee\":\"2\",\"currency\":\"EUR\",
+      \"asset\":{\"symbol\":\"$symbol\",\"quoteCcy\":\"EUR\"}}" >/dev/null
+  done
+done
+for month in 03 06 09 12; do
+  activity "{\"accountId\":\"$BROKER\",\"activityType\":\"DIVIDEND\",\"activityDate\":\"2025-$month-20\",
+    \"amount\":\"12.5\",\"currency\":\"USD\",\"asset\":{\"symbol\":\"AAPL\"}}" >/dev/null
+done
+echo "investments: done" >&2
+
+# Spending: the account must be tracked before its activities count.
+wf call update_spending_settings --args \
+  "{\"update\":{\"enabled\":true,\"accountIds\":[\"$CHECKING\"]}}" >/dev/null
+categorise() {
+  wf call assign_activity_category --args \
+    "{\"activityId\":\"$1\",\"taxonomyId\":\"$2\",\"categoryId\":\"$3\"}" >/dev/null
+}
+salary=$(activity "{\"accountId\":\"$CHECKING\",\"activityType\":\"DEPOSIT\",\"activityDate\":\"2025-11-01\",\"amount\":\"6500\",\"currency\":\"USD\"}")
+categorise "$salary" income_sources cat_income_employment
+# day category amount
+while read -r day category amount; do
+  spend=$(activity "{\"accountId\":\"$CHECKING\",\"activityType\":\"WITHDRAWAL\",
+    \"activityDate\":\"2025-11-$day\",\"amount\":\"$amount\",\"currency\":\"USD\"}")
+  categorise "$spend" spending_categories "$category"
+done <<'ROWS'
+01 cat_housing 2200
+02 cat_groceries 84.20
+03 cat_transport 45
+04 cat_food 32.50
+05 cat_bills 120
+06 cat_groceries 61.75
+07 cat_entertainment 18.99
+08 cat_shopping 149
+09 cat_food 27.80
+10 cat_health 60
+11 cat_groceries 92.10
+12 cat_transport 45
+13 cat_food 41.20
+14 cat_personal 35
+15 cat_bills 64.99
+16 cat_groceries 73.40
+17 cat_travel 410
+18 cat_food 22.30
+19 cat_gifts 75
+20 cat_groceries 88.60
+21 cat_entertainment 12.99
+22 cat_fees 4.50
+23 cat_food 55.10
+24 cat_shopping 89
+25 cat_groceries 66.30
+26 cat_transport 45
+27 cat_education 120
+28 cat_food 38.90
+29 cat_other_expense 25
+30 cat_groceries 79.95
+ROWS
+while read -r category amount; do
+  wf call upsert_budget_target --args "{\"periodKey\":\"2025-11\",\"target\":{\"periodKey\":\"2025-11\",
+    \"targetType\":\"category\",\"taxonomyId\":\"spending_categories\",\"categoryId\":\"$category\",
+    \"amount\":\"$amount\"}}" >/dev/null
+done <<'ROWS'
+cat_housing 2200
+cat_groceries 500
+cat_food 200
+cat_transport 150
+cat_bills 200
+cat_shopping 150
+cat_entertainment 50
+ROWS
+echo "spending: done" >&2
+
+# Quotes and the recalculation; the sync result says which symbols failed.
+wf sync --wait --wait-timeout 300 >&2

scripts/bench.sh

@@ -0,0 +1,91 @@
+#!/usr/bin/env bash
+# Measures how much each read command prints, as a Markdown table: what an
+# agent pays to read it. Tokens are estimated as bytes / 4.
+#
+#   scripts/bench.sh            # dist/wf.js against the seeded server
+#   scripts/bench.sh <git-ref>  # also builds <git-ref> and compares
+#
+# Needs a server seeded by scripts/bench-seed.sh (it reuses its login), and a
+# build (pnpm build). Runs only reads, so it can run any number of times.
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+export WF_CONFIG_DIR="${WF_BENCH_CONFIG_DIR:-$ROOT/.bench}"
+NEW="$ROOT/dist/wf.js"
+[ -f "$NEW" ] || { echo "Run pnpm build first." >&2; exit 2; }
+[ -d "$WF_CONFIG_DIR" ] || { echo "Run scripts/bench-seed.sh first." >&2; exit 2; }
+
+BASE=""
+if [ -n "${1:-}" ]; then
+  TREE="$(mktemp -d)"
+  trap 'git -C "$ROOT" worktree remove --force "$TREE" >/dev/null 2>&1 || true' EXIT
+  git -C "$ROOT" worktree add -q --detach "$TREE" "$1"
+  ln -s "$ROOT/node_modules" "$TREE/node_modules"
+  (cd "$TREE" && pnpm -s build >/dev/null)
+  BASE="$TREE/dist/wf.js"
+fi
+
+account() {
+  node "$NEW" accounts | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{
+    const a=JSON.parse(s).find((x)=>x.name===process.argv[1]); console.log(a ? a.id : "")})' "$1"
+}
+BROKER="$(account Brokerage)"
+[ -n "$BROKER" ] || { echo "No Brokerage account. Run scripts/bench-seed.sh first." >&2; exit 2; }
+
+# label|arguments. <broker> is replaced by the Brokerage account id.
+COMMANDS=(
+  "status|status"
+  "accounts|accounts"
+  "holdings|holdings"
+  "holdings --account|holdings --account <broker>"
+  "holdings --full|holdings --full"
+  "activity search|activity search"
+  "activity search --page-size 100|activity search --page-size 100"
+  "activity search --full|activity search --full"
+  "health|health"
+  "budget|budget --period 2025-11"
+  "budget --full|budget --period 2025-11 --full"
+  "spending report|spending report --from 2025-11-01 --to 2025-11-30"
+  "spending report --full|spending report --from 2025-11-01 --to 2025-11-30 --full"
+  "commands|commands"
+  "schema create_activity|schema create_activity"
+  "--help|--help"
+)
+
+# Bytes on stdout, or "exit N" when the command fails.
+measure() {
+  local bin=$1 out code
+  shift
+  set +e
+  out="$(node "$bin" "$@" 2>/dev/null | wc -c)"
+  code=${PIPESTATUS[0]}
+  set -e
+  if [ "$code" != 0 ]; then echo "exit $code"; else echo "$out"; fi
+}
+
+cell() { case $1 in exit*) echo "$1" ;; *) echo "$1 (~$(($1 / 4)))" ;; esac; }
+
+if [ -n "$BASE" ]; then
+  echo "| Command | $1 bytes (~tokens) | dist bytes (~tokens) | Change |"
+  echo "| --- | ---: | ---: | ---: |"
+else
+  echo "| Command | Bytes (~tokens) |"
+  echo "| --- | ---: |"
+fi
+for entry in "${COMMANDS[@]}"; do
+  label="${entry%%|*}"
+  read -r -a args <<<"${entry#*|}"
+  args=("${args[@]/<broker>/$BROKER}")
+  now="$(measure "$NEW" "${args[@]}")"
+  if [ -n "$BASE" ]; then
+    before="$(measure "$BASE" "${args[@]}")"
+    change=""
+    case "$before$now" in
+      *exit*) ;;
+      *) [ "$before" -gt 0 ] && change="$(((now - before) * 100 / before))%" ;;
+    esac
+    echo "| \`$label\` | $(cell "$before") | $(cell "$now") | $change |"
+  else
+    echo "| \`$label\` | $(cell "$now") |"
+  fi
+done

scripts/test-server.sh

@@ -6,6 +6,9 @@
 #
 # The image tag comes from UPSTREAM_TAG without the leading "v".
 # WF_TEST_PORT (default 8088) and WF_TEST_TTL_MINUTES (default 60) are optional.
+# WF_TEST_PROXY=1 passes HTTPS_PROXY and WF_TEST_CA_BUNDLE (a CA file) to the
+# server, for sandboxes where outbound traffic must go through a proxy, so
+# quote syncs can reach the market data provider.
 set -euo pipefail
 
 ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
@@ -20,8 +23,18 @@ HASH='$argon2id$v=19$m=65536,t=3,p=4$MmjLEewIk4DVBqvLmF5+Dg$0yHuiPF4d7YOOS9+tnOV
 case "${1:-}" in
   up)
     docker rm -f "$NAME" >/dev/null 2>&1 || true
-    docker run -d --name "$NAME" -p "127.0.0.1:$PORT:8088" \
-      -e WF_LISTEN_ADDR=0.0.0.0:8088 \
+    NET=(-p "127.0.0.1:$PORT:8088" -e WF_LISTEN_ADDR=0.0.0.0:8088)
+    if [ "${WF_TEST_PROXY:-}" = 1 ]; then
+      # A proxy on the host's loopback is reachable only from the host network.
+      NET=(--network host -e "WF_LISTEN_ADDR=127.0.0.1:$PORT"
+        -e "HTTPS_PROXY=${HTTPS_PROXY:?WF_TEST_PROXY=1 needs HTTPS_PROXY}"
+        -e "HTTP_PROXY=${HTTP_PROXY:-$HTTPS_PROXY}" -e NO_PROXY=127.0.0.1,localhost)
+      if [ -n "${WF_TEST_CA_BUNDLE:-}" ]; then
+        NET+=(-v "$WF_TEST_CA_BUNDLE:/etc/ssl/certs/ca-certificates.crt:ro"
+          -e SSL_CERT_FILE=/etc/ssl/certs/ca-certificates.crt)
+      fi
+    fi
+    docker run -d --name "$NAME" "${NET[@]}" \
       -e WF_DB_PATH=/data/wealthfolio.db \
       -e "WF_SECRET_KEY=$(openssl rand -base64 32)" \
       -e "WF_AUTH_PASSWORD_HASH=$HASH" \

skills/wealthfolio/SKILL.md

@@ -25,8 +25,8 @@ code you can act on.
    `wf activity search`. Never guess an id.
 5. **Add `--wait` to writes when you read right after.** Without it, the
    server recalculates in the background and your next read can be stale.
-6. **Read as JSON, without `--table`.** `--table` truncates nested fields such
-   as `instrument`, so you lose the symbol. Fetch JSON, then summarise or
+6. **Read as JSON, without `--table`.** `--table` cuts long values and nested
+   fields, so you can lose data. Fetch JSON, then summarise or
    format it yourself, even when the user asks to see the data. Use `--table`
    only when the user asks for a table.
 
@@ -77,18 +77,15 @@ wf rebalance --cash 5000             # plus a plan to invest 5000
   round them.
 - Dates are `YYYY-MM-DD` or RFC 3339.
 - Add `--table` only when the user asks for a table (rule 6).
-- `wf holdings` leaves out each instrument's classifications and notes and the
-  holding's metadata. Add `--full` only when you need sectors or regions. Even
-  slim, no `--account` returns the aggregate across all accounts (on one
-  portfolio, 26 rows were about 12k tokens), so pass `--account` when the question is about one
-  account. Pipe it through `jq` to the fields you need:
-
-  ```sh
-  wf holdings --account <id> | jq '[.[] | {symbol: .instrument.symbol,
-    quantity, price, value: .marketValue.base, cost: .costBasis.base,
-    gain: .totalGain.base, weight}]'
-  ```
-
+- `wf holdings` returns one flat row per holding: `assetId`, `symbol`,
+  `quantity`, `price` (in `currency`), and `marketValue`, `costBasis`,
+  `unrealizedGain`, `totalGain`, `totalGainPct`, `dayChangePct` and `weight`
+  in `baseCurrency`. Null fields are left out. Add `--full` only when you need
+  local values, lots, sectors or regions. Without `--account`, it returns the
+  total across all accounts, with `sourceAccountIds`. Pass `--account` when the
+  question is about one account.
+- Add `--fields a,b` to any command to keep only those fields of each row,
+  for example `wf holdings --fields symbol,marketValue,weight`.
 - `activity search` returns 20 rows by default, each with only `id`, `accountId`,
   `assetId`, `assetSymbol`, `activityType`, `subtype`, `status`, `date`,
   `quantity`, `unitPrice`, `amount`, `fee`, `currency`, `fxRate` and
@@ -100,12 +97,15 @@ wf rebalance --cash 5000             # plus a plan to invest 5000
   it was recorded at.
 - `wf spending report` defaults to this month. It returns `current` and
   `prior` totals (`income`, `outflow`, `saved`, `net`) and breakdowns by
-  category and day. `--account <id>` repeats.
-- `wf budget` returns `state` (groups, targets) and `computed` (`totals`,
-  per-group rows). All zeros with empty `groupRows` means no budget is set up
-  for that month; say so instead of reporting zero spending.
+  category. Add `--full` for the per-day totals. `--account <id>` repeats.
+- `wf budget` returns `totals`, `groups` (each with `planned`, `actual`,
+  `remaining`, `overspent` and its `categories`), `ungrouped` and `income`.
+  Categories with no target and no spending are left out. Zero totals with no
+  categories mean no budget is set up for that month; say so instead of
+  reporting zero spending. `--full` returns the server's `{state, computed}`.
 - `wf health` leaves out each issue's `diagnostics` (it repeats `details` and
-  `affectedItems`). Add `--full` only if those don't explain an issue.
+  `affectedItems`) and the web app's routes. Add `--full` only if those don't
+  explain an issue.
 - `wf holdings --account <id> --include-closed` also lists sold-out positions.
 
 For anything without its own command:
@@ -131,7 +131,7 @@ wf call create_activity --wait --args '{"activity":{
 
 - A new asset needs `symbol` and `quoteCcy`. For an asset already stored,
   pass `"asset":{"id":"<assetId>"}`. `wf holdings` shows that id as
-  `instrument.id`.
+  `assetId`.
 
 Cash activities (`DEPOSIT`, `WITHDRAWAL`, `DIVIDEND`, `INTEREST`, `FEE`) take
 `amount` and no quantity. The output is `{result, wait}`. When
@@ -161,7 +161,7 @@ wf activity update <id> --unit-price 110 --wait
 - Name only the fields that change. `--clear fee` sets an optional field to
   null.
 - `--asset` takes an asset id, not a symbol. Find the id in `wf holdings`
-  (`instrument.id`) or in the activity's `assetId`.
+  or in the activity, as `assetId`.
 - `--approve` takes an activity out of the review queue. A draft becomes
   posted.
 - For a batch, use `--updates '[{"activityId":"a","fee":"1"},...]'` or