flow) and cross-symbol signed-bar comparison (cohort) — both Recharts —\nplus SEPA strategy dashboard (sepa) and short-term multi-timeframe\nprediction dashboard (intraday) — both TradingView Lightweight Charts.\nMulti-timeframe K-line review lives inside intraday (the standalone kline\nchart type was removed). The server fetches Longbridge data itself\n(kline / capital flow) and computes all indicators (MA, MACD, RS, trend\ntemplate, volume profile, divergence/beichi detection) in TypeScript; the\ncaller only POSTs {type, symbol, ...} to /api/charts and gets back\n{id, url, technicals?}. Charts persist as data JSON under\njournal/charts/data/. url now points at where the chart actually lives:\nsepa/intraday land on the symbol page (http://localhost:1792/symbol/<SYM>?analysis=<id>),\nflow/cohort land on the home page for that date (http://localhost:1792/?date=YYYY-MM-DD);\nold http://localhost:1792/charts/<id> links still work and redirect there.\nTriggers: 出图、生成图表、画 K 线、画资金流曲线、画对比图、SEPA 仪表盘、\n短线预测、多周期K线、MACD、入场判断可视化、可视化、render chart, plot,\nvisualise, sepa dashboard, intraday prediction dashboard.Creates charts through the local chart app so the user can browse them in one\nplace instead of squinting at tables. The server pulls market data and computes\neverything; charts are stored as versioned data JSON and always rendered by the\nlatest front-end code.
\n\n\nResponse language: match the user — 简体 / 繁體 / English.
\n
longbridge capital --flow context or when the user wants a flow visual ⇒ flowintradaycohortsepa-strategy on a single name ⇒ sepaintraday-signal ⇒ intraday (two-call pattern: POST preview → PATCH prediction)capital-rotation / market-session-tracker / stock-deep-dive,\ncall this as the LAST step and append the chart URL to the markdown journal entry.Skip when the user only wants a single number or a tiny series — a Unicode\nsparkline in the chat reply is faster.
\n在 Kansoku 数据目录里跑（外部 Claude Code、无本地 dev server），走\nkansoku-cli 而不是 HTTP。判据：环境变量 $KANSOKU_CLI 有值 → 走 CLI；\n无 → 回退 HTTP。
CLI 与 HTTP 契约一致——body 就是原本 POST /api/charts 的 body，\n返回 {id, url, technicals?} 也不变；CLI 侧额外多带一个 deepLink\n字段（见下方数据目录场景说明）。
先加载 runtime.env（这一步只需做一次）：
\nset -a\n[ -f .kansoku-agent-kit/runtime.env ] && . .kansoku-agent-kit/runtime.env\nset +a\n之后：
\n| HTTP 版 | \nCLI 等价 | \n
|---|---|
curl -s http://localhost:1792/api/health | \n\"$KANSOKU_CLI\" info kit-version | \n
POST /api/charts body 为 X | \nprintf '%s' \"$X\" \\| \"$KANSOKU_CLI\" chart create --type <T> --symbol <S> --json-input - | \n
GET /api/charts?type=&symbol= | \n\"$KANSOKU_CLI\" chart list [--symbol X] | \n
GET /api/charts/<id> | \n\"$KANSOKU_CLI\" chart get <id> | \n
PATCH /api/charts/<id> | \n(暂不支持——PATCH 仍走 HTTP；数据目录场景 PATCH 目前只在 intraday-signal 里出现，届时 fallback HTTP) | \n
打开 data.url（浏览器访问 http://localhost:1792/...） | \n打开 data.deepLink（kansoku://route/...，见下） | \n
body/response 完全一致，写入的图表 JSON 落在 journal/charts/data/<id>.json\n（跟 HTTP 版本相同）。
数据目录场景通常没有本机 dev server（Kit 用户不会跑 pnpm start），\ndata.url 里的 http://localhost:1792/... 打不开。CLI 的 chart create\n响应现在多带一个 deepLink 字段（url 照旧保留，向后兼容 dev 场景），\n指向 kansoku://route/... 这个自定义协议。想让用户看到图表时，打开这个\ndeep link 即可——它会启动 Kansoku.app（若已在运行则直接切到前台）并跳转到\n对应图表页面：
open \"$(echo \"$response\" | jq -r .deepLink)\"\nmacOS 用 open，Windows 用 start，Linux 用 xdg-open——三者都认协议处理器，\n换平台只需换这一条命令。
The app must be running before any API call:
\ncurl -s http://localhost:1792/api/health # {\"ok\":true,...} = up\nCLI 等价：\"$KANSOKU_CLI\" info kit-version
If it is down, start it (long-running process — use run_in_background):
\npnpm start # serves API + built web UI on :1792\nFirst-time setup only: pnpm install && pnpm build.
Base URL http://localhost:1792. All responses follow the\n{ok, data, meta} / {ok:false, error, hint} contract.
| Endpoint | \nPurpose | \n
|---|---|
GET /api/health | \nliveness check | \n
GET /api/charts?type=&symbol=&limit= | \nlist chart metas (newest first) | \n
POST /api/charts | \ncreate a chart; body below | \n
GET /api/charts/:id | \nfull chart doc | \n
GET /api/charts/:id/built?count= | \nephemeral intraday rebuild with a larger bar window (history view; max 1000, never persisted) | \n
PATCH /api/charts/:id | \nmerge fields into input and rebuild (e.g. add prediction) | \n
DELETE /api/charts/:id | \nremove a chart | \n
GET /api/legacy | \nlist old single-file HTML archives (served at /legacy/<file>) | \n
GET /api/stream/quotes?extra= | \nSSE quote snapshots (watchlist ∪ positions ∪ extra), 10s cadence | \n
GET /api/stream/charts/:id | \nSSE live rebuilds for flow/intraday charts, 60s cadence | \n
GET /api/symbols/:sym/{flow,benchmark,position,analyses,latest} | \nlive per-symbol cockpit data (server-computed, never AI) | \n
The stream endpoints power the web UI's realtime display; the AI workflow never\nneeds them — created charts update themselves in the browser while open, and\nthe persisted JSON stays frozen at analysis time.
\n/symbol/<SYM>)Every symbol also gets a stable dashboard URL, http://localhost:1792/symbol/<SYM>,\nthat aggregates live data with the symbol's latest intraday/sepa analysis — it\nis the caller-facing counterpart of GET /api/symbols/:sym/*. The same page also\nserves the frozen per-analysis view: appending ?analysis=<id> pins it to one\nspecific chart doc (this is what data.url returns for sepa/intraday charts)\ninstead of following the latest one live; a banner offers to jump back to latest\nif a newer analysis has since landed. Old /charts/<id> links still work — they\nresolve the chart's type client-side and redirect to this page (or to the home\npage for flow/cohort charts). During regular session the cockpit also carries a live AI comment stream and auto-reassessment (commentator + escalated analyst, gated on the comment/analyst model settings configured at /settings) whose output shares the manual intraday-signal format. These /api/symbols/*\nroutes are server plumbing for that page — direct callers rarely need them,\nlisted here only for completeness:
GET /api/symbols/:sym/flow — today's capital-flow curve + large/medium/small distribution (live, via longbridge capital)GET /api/symbols/:sym/benchmark — SMH/QQQ normalized same-session comparison (live, via longbridge kline)GET /api/symbols/:sym/position — shares/cost/unrealized + distance to stop/target from the latest analysis's entry plan (live)GET /api/symbols/:sym/analyses — past intraday analyses for this symbol with mechanical outcome judgments (hit_target / hit_stop / open, computed server-side from post-anchor bars — never AI recall)GET /api/symbols/:sym/latest — the latest intraday chart doc in full, plus prediction_staleThe client-side indicator toggle bar (show/hide 金叉死叉、自动背离、自动背驰、123\n结构、SB 结构、K线形态、AI 标注、价位线、EMA 均线; state in localStorage) has no API\nsurface — it's a pure front-end feature on both the cockpit and archive pages.\nSwing 高低点 are not one of the toggles — they're baked into the\ndivergence/beichi/pattern123/SB detectors as the underlying pivot data, not a\nstandalone overlay (deliberate deviation from the earlier spec draft). SB（第二次\n突破/跌破，对应 Al Brooks 的 High 2 / Low 2）标注顺势回调后第二次尝试冲破前高/前低\n的结构：图上灰色 H1/L1 是失败的第一次尝试，金色 H2/L2 箭头是确认成功的第二次突破；\ntechnicals 里对应 second_breakouts 这一行。
The server fetches Longbridge data itself when symbol is given; pass data\n(or kline / timeframes) only to override with hand-assembled rows.
// flow — server runs `longbridge capital <SYM> --flow`\n{ \"type\": \"flow\", \"symbol\": \"MU.US\", \"subtitle\": \"单位推断为千 USD · 仅供参考\" }\n\n// cohort — data is always caller-assembled\n{ \"type\": \"cohort\", \"title\": \"存储 vs Mag 7 主力净流\",\n \"data\": [{ \"symbol\": \"MU\", \"value\": -17087 }, { \"symbol\": \"NVDA\", \"value\": 9540 }] }\n\n// sepa — server pulls 260 day bars + SPY.US automatically\n{ \"type\": \"sepa\", \"symbol\": \"MRVL.US\", \"name\": \"Marvell Technology\",\n \"position\": { \"shares\": 1, \"cost\": 100.00 }, // optional\n \"context\": { /* see sepa context schema below */ } }\n\n// intraday — server pulls 5m/15m/1h × 1000 bars automatically (--session all:\n// pre/post-market bars included by default; pass \"session\": \"intraday\" to exclude)\n{ \"type\": \"intraday\", \"symbol\": \"MU.US\", \"name\": \"Micron Technology\",\n \"ema_periods\": [9, 21, 55], // optional, fast/mid/slow EMA overlay (default 9/21/55, max 4)\n \"position\": { \"shares\": 1, \"cost\": 100.00 }, // optional\n \"prediction\": null } // omit for preview mode\nCLI 等价（body 走 stdin，--json-input -）：
\"$KANSOKU_CLI\" chart create --type flow --symbol MU.US --json-input -\"$KANSOKU_CLI\" chart create --type cohort --json-input -（无单一 symbol，省略 --symbol）\"$KANSOKU_CLI\" chart create --type sepa --symbol MRVL.US --json-input -\"$KANSOKU_CLI\" chart create --type intraday --symbol MU.US --json-input -Success returns data.id, data.url (paste this into journal entries), plus\ntype-specific meta: sepa → verdict_tier / passes / fails / bars; intraday →\nmode / bars / technicals.
context schemaAll fields optional:
\n{\n \"earnings_dates\": [\"2026-05-29\"], // E markers on those bars\n \"stage\": \"Stage 2 末期\", // 阶段判断 sidebar card\n \"stage_note\": \"Stage 3 顶部嫌疑\",\n \"base_count\": \"3-4 (减半仓)\",\n \"pattern\": \"无可买（扩张振幅）\",\n \"verdict\": {\n // override the auto verdict\n \"tier\": \"watch\", // pass / watch / buy\n \"label\": \"\uD83D\uDC40 WATCH LIST\",\n \"color\": \"#ffc107\",\n \"reason\": \"...\",\n },\n \"entry_plan\": {\n // 入场计划 card + price lines\n \"pivot\": 260.0, // required: consolidation-range high\n \"stop\": 241.8, // default pivot × 0.93 (-7%)\n \"target1_pct\": 8, // default 8 (Phase 2: 卖一半 + 移至本钱)\n \"target2_pct\": 15, // default 15 (Phase 3: 再卖 25% + 沿 20MA 跟踪)\n \"note\": \"...\",\n \"hypothetical\": true, // 标注 \"假设性\" 徽章\n },\n \"support_zones\": [\n // omit → auto zones (MA50 / MA200 / volume cluster)\n {\n \"low\": 217,\n \"high\": 226,\n \"tier\": \"watch\", // warning / watch / buy / value\n \"label\": \"MA50 关注区\",\n \"note\": \"...\",\n \"sources\": [\"MA50 $221.75\"],\n },\n ],\n \"auto_support_zones\": true, // false disables the auto fallback\n \"volume_profile\": { \"lookback_days\": 120, \"bins\": 30 },\n}\nDerived values (auto-computed server-side): buy_zone_high = pivot × 1.05,\ntarget1/2 = pivot × (1 + pct/100), R/R = (target2 − pivot) / (pivot − stop)\n— based on T2, not T1, because T1 is the SEPA Phase-2 partial exit. R/R < 2:1\nrenders a red warning.
Verdict auto-detection (when context.verdict omitted): any trend-template\nfail → PASS \uD83D\uDEAB; all 8 pass + price ≥ 25% above MA50 → WATCH · Extended \uD83D\uDC40;\nall 8 pass otherwise → WATCH · No pattern detected \uD83D\uDC40. STRONG BUY ✅ is never\nauto-emitted — pass context.verdict after manually confirming a valid pattern
Auto markers on the main K-line: earnings (context.earnings_dates),\nclimax top (volume ≥ 2.5×20MA + red close + local high), MA50/MA200 breakdowns,\n52w high. Hardcoded lines: 52w high/low, MA50 × 1.25 extended warning.
prediction → preview. Read data.technicals from the\nresponse: per timeframe last_dif / last_dea / last_hist, last_vwap\n(当日 session VWAP，m5/m15), emas (latest\nfast/mid/slow EMA values — price vs EMA stack tells the short-term trend\nposture), recent swing highs/lows, last_cross (金叉/死叉),\ndivergence_candidates, beichi_candidates, second_breakouts (SB 结构，\nH2/L2). Also read meta.day_context:\ndaily_trend (up/down/range vs 日线 MA20/MA50), daily_ma20/daily_ma50,\nhigh_20d/low_20d, prev_day (昨日高/低/收), pre_market 区间,\nopening_range (开盘前 30 分钟), vwap — the server draws these as the\nVWAP line and 日内参照位 price lines on the chart. Read these numbers —\ndon't eyeball candles./api/charts/:id with {\"prediction\": {...}} → final dashboard.\nAdd \"refresh\": true to any PATCH to refetch the latest bars (incl. pre/post\nmarket) and recompute everything before rebuilding — same id, same URL.CLI 等价：第 1 步（POST 预览）可用\n\"$KANSOKU_CLI\" chart create --type intraday --symbol MU.US --json-input -；\n第 2 步（PATCH 写入 prediction）目前无 CLI 等价，仍走 HTTP。
prediction schema:
{\n \"direction\": \"short\", // long | short | neutral\n \"anchor\": { \"timeframe\": \"m15\", \"time\": \"2026-07-01T17:00:00Z\", \"price\": 1049.81 },\n \"scenarios\": [\n // ≥ 2, probabilities ≈ 100\n { \"label\": \"继续探底\", \"probability\": 45, \"path\": \"...\", \"trigger\": \"...\" },\n ],\n \"range_bound_plan\": {\n \"condition\": \"...\",\n \"long_tactic\": \"...\",\n \"short_tactic\": \"...\",\n \"low\": 1020.0,\n \"high\": 1060.0,\n },\n // low/high = 箱体下沿/上沿；neutral 必填（观望按守住/破位对账入记分板），方向单可选\n \"entry_plan\": {\n \"entry\": 1049.81,\n \"stop\": 1030.0,\n \"target1_pct\": 3,\n \"target2_pct\": 6,\n \"note\": \"...\",\n }, // 仅 long/short；neutral（观望）省略整个 entry_plan，条件应对写进 range_bound_plan\n \"price_zones\": [\n // only real resistance/pressure zones drawn on chart\n {\n \"kind\": \"resistance\",\n \"label\": \"反弹压力带\",\n \"low\": 60.9,\n \"high\": 61.35,\n \"note\": \"短线均线和第一修复位重合\",\n \"sources\": [\"5m EMA9/21\", \"第一修复位\"],\n },\n ],\n \"signals\": [\n // 可选；背离/背驰/K线形态/123 结构/SB 结构均由服务端自动检测绘制，无需在此重复\n {\n \"type\": \"other\",\n \"timeframe\": \"m5\",\n \"time\": \"...\",\n \"price\": 1032.28,\n \"bias\": \"bearish\",\n \"label\": \"尾盘放量长阴收在最低——摆动点未确认前的人工备注\",\n },\n ],\n}\nR/R is direction-aware (long: risk = entry−stop; short: risk = stop−entry);\nthe sidebar flags rr < 2:1 in red. entry_plan can carry structured level\ncontext so the chart explains why a point was selected instead of hiding the\nreason in prose:
{\n \"entry\": 61.1,\n \"stop\": 62.52,\n \"target1\": 60.0, // optional explicit target price; overrides pct-derived price\n \"target2\": 57.92,\n \"rationale\": \"反弹到 60.90-61.35 压力带后受阻才入场。\",\n \"stop_note\": \"站回上一段反弹高点，空头计划失效。\",\n \"entry_zone\": { \"kind\": \"resistance\", \"label\": \"反弹压力带\", \"low\": 60.9, \"high\": 61.35 },\n \"target1_label\": \"T1 · 日内低点\",\n \"target1_note\": \"整数位和日内低点，首次触及先看是否止跌。\",\n \"target1_zone\": { \"kind\": \"support\", \"label\": \"日内低点\", \"low\": 60.0, \"high\": 60.0 },\n \"target2_label\": \"T2 · 深一档支撑\",\n \"target2_condition\": \"60.00 跌破并反抽失败后才成立。\",\n \"target2_zone\": { \"kind\": \"support\", \"label\": \"深一档支撑\", \"low\": 57.9, \"high\": 58.0 },\n}\nentry_plan.entry_zone and target1_zone / target2_zone are explanation\ncontext for the right panel only; they are not drawn as chart zones and should\nnot be named 入场区, T1 区域, or T2 区域. Put only genuine upper supply\nareas in top-level price_zones with kind: \"resistance\"; those are rendered\nas chart boundaries and in the sidebar's key-zone section.
Supported zone kinds remain entry, stop, target, support, resistance,\ninvalidation, watch, but the intraday chart-zone overlay intentionally\nfilters to explicit resistance zones. MACD structure signals + simplified 背离/背驰\nare auto-detected and drawn on every render regardless of prediction. Every\nDIF/DEA cross is classified by zero-line position（零上/零下金叉、零上/零下死叉）\nplus structural patterns: 二次金叉（零下双金叉且低点抬高 → 底部确认）、空中加油\n（零上二次金叉 → 强势延续）、二次死叉（顶部确认 / 空头中继）、上穿/下穿零轴\n（趋势确认，last 2 bars marked 待确认 with a ?）. Each marker's tooltip carries\nthe implication text; technicals.<tf>.structure_signals (last 6) and\nzero_tangle（DIF 贴零轴缠绕 = 震荡市，交叉信号失效）expose the same data to the\nanalysis workflow — read them in the preview step. The swing-based divergence\nauto-detectors only fire on confirmed swing pivots — the last 1-2 bars can never\nbe flagged; read last_dif/last_dea/last_hist directly for the newest bar.
14 classic K-line patterns are also auto-detected on the main pane (arrow\nmarkers, tooltip carries the implication): 单根——锤子线/上吊线/倒锤子/射击之星;\n双根——看涨吞没/看跌吞没/乌云盖顶/刺透形态/看涨孕线/看跌孕线; 三根——启明星/\n黄昏星/红三兵/三只乌鸦. All require a trend-context filter (4-bar net move into\nthe pattern) plus a body-size filter vs the trailing 14-bar average, so quiet\nchop produces few marks. One pattern max per bar (stars > soldiers/crows >\ntwo-bar > single-bar). technicals.<tf>.candle_patterns (last 6) exposes them\nto the analysis workflow. Caveat: the newest bar may still be forming intraday —\na pattern on it can repaint until the bar closes.
123 reversal structures (Sperandeo 1-2-3) are also auto-detected per timeframe\nfrom confirmed swing pivots: ① a ~20-bar extreme → ② the reaction pivot → ③ a\nhigher low (bullish) / lower high (bearish). The price pane gets ①②③ markers, a\n①→②→③ connector, and a dashed trigger line at the ② price running from ③ to the\nlatest bar; a close beyond ② flips the structure to confirmed (123✓ marker on\nthe breakout bar), while a break of ① drops the structure silently. The ③ marker\nshows ③? until confirmation. technicals.<tf>.pattern_123 (last 2, each with\nstatus: forming|confirmed, trigger, invalidation, p1/p2/p3) exposes them\nto the analysis workflow — a forming structure is a ready-made entry setup\n(enter on the ② break, stop beyond ①). The sidebar's 自动信号 section lists them\nwith an 酝酿中/已确认 badge.
Off-session bars (盘前/盘后浅蓝、夜盘深蓝) get a full-height backdrop on both\npanes — thin-volume price action outside regular hours is visually discounted at\na glance. Regular hours = 09:30-16:00 ET (DST-aware via America/New_York).
\ncontext — AI-classified news + conclusion (optional, schema_version 2)Both POST /api/charts (type intraday) and PATCH /api/charts/:id accept an\noptional context field alongside prediction. It's frozen at write time like\nprediction — the server never generates or judges it. schema_version is now\n2; older (v1) chart docs without context still load and render fine.
\"context\": {\n \"generated_at\": \"2026-07-06T14:30:00Z\", // ISO timestamp\n \"conclusion\": {\n \"stance\": \"short\", // long | short | neutral\n \"summary\": \"一句话综合判断\",\n \"action\": \"现在该做什么（挂单/等待/减仓）\"\n },\n \"news\": [\n { \"time\": \"2026-07-06T13:10:00Z\",\n \"source\": \"longbridge\", // longbridge | x | trump | sec | gdelt\n \"tag\": \"catalyst\", // catalyst | regulatory | sentiment | macro\n \"title\": \"...\", \"note\": \"AI 一句话解读\", \"url\": \"可选\" }\n ],\n \"sources_used\": [\"longbridge-news\", \"twitter-reader\"]\n}\nThe dashboard shows context.generated_at's age and a stale badge, sharing the\nsame ~15-min staleness rule as prediction (prediction_updated_at /\nprediction_stale on chart metas cover both).
Once the US cash session is open, intraday charts must be maintained under\nthese rules:
\n{\"session\": \"intraday\", \"refresh\": true} — the prediction\ndashboard must not render pre-market / overnight bars intraday. The\ndefault --session all is for pre-market analysis only.longbridge kline --period 5m (regular-session bars), sum today's bars, and compare against\nthe same number of opening bars averaged over the prior ~5-8 sessions.\nPair the ratio with per-bar direction before calling a move confirmed — a\nlevel break on ~0.6x same-period volume is not a confirmed breakout.\nCaveat: Longbridge daily-K volume includes extended hours; 5m-K volume\ndoes not — never mix the two.GET /api/charts?stale=true lists them; the SSE envelope and chart\nmetas carry prediction_updated_at / prediction_stale). Each loop\nround: fetch the stale list → re-pull quote / capital flow / klines →\nPATCH prediction with scenarios revised only on material change, but\nalways move anchor to the newest m5 bar time + latest price — the\nanchor marker must track the live tape, never sit minutes behind it →\nappend a timestamped journal note on material revisions (revision\ndiscipline) → stop the loop after 16:00 ET close.打包的桌面版（Kansoku.app）不监听任何本机端口，全部走 app:// 内部协议 + IPC，不提供 HTTP 接口。要用这个 skill 的 curl 命令，单独起 server 进程即可：pnpm start（或开发模式 pnpm dev），API 照常在 http://localhost:1792。
journal/charts/data/<YYYY-MM-DD>-<slug>.json — gitignored,\nschema_version field for forward compatibility. Date = US session date\n(derived from the data, not local clock).journal/charts/*.html, listed in the\napp under 旧版存档 and served at /legacy/<file>.apps/server/ Hono +\nTS, apps/web/ Vite + React). Analysis parity with the retired Python\nimplementation is locked by vitest golden tests: pnpm test.For tiny in-chat previews render Unicode sparklines directly: ▁▂▄▆█ plus ANSI\ngreen/red. Use for 5-20-point series where a full chart is overkill.
longbridge-capital-flow / longbridge-kline — same data the server pulls; call directly only for in-chat analysiscapital-rotation — should end with a cohort chartmarket-session-tracker — may create flow chartssepa-strategy — calls sepa as the last step of its Step 10intraday-signal — calls intraday twice (POST preview, then PATCH prediction)