某消费股 · 食品安全事件 Consumer stock · food safety
- Douyin+0.62
- 公众号−0.74
- CLS−0.51
提前 30 分钟触发一致性预警,避开次日开盘 −7% 缺口。 Pre-warning fired 30 min before open. Avoided the next-day −7% gap.
内测开放中 · 限额 100 席 Private beta · 100 seats
不是再做信息聚合。alphapulse 当前已接通 4 个核心平台(财联社 / 公众号 / 抖音 / YouTube),先把 情绪 → 一致性 → 因子信号这条链路收回来。 系统输出来源一致性评分、冲突标签和可复盘的结构化信号,三 Agent 并发后直接给研究系统取数。 Stop aggregating feeds by hand. alphapulse is live on 4 core platforms today (CLS, WeChat MP, Douyin, YouTube). It turns sentiment into a cross-market factor with consistency scoring, conflict labeling, and structured outputs for your strategy stack.
语音和文本都是输入通道。完整研究流水线需启动 :10510 并设置 VOICE_API_BASE,返回可复盘结构化因子。
Voice and text are input channels. Enable the full chain on :10510 with VOICE_API_BASE to return auditable factor output.
01 — 信号噪音01 — Signal vs. noise
提前 30 分钟触发一致性预警,避开次日开盘 −7% 缺口。 Pre-warning fired 30 min before open. Avoided the next-day −7% gap.
四平台高度共振,研究员当日纳入产业链跟踪池。 4 platforms resonating. Added to supply-chain tracking pool same day.
02 — 系统能力02 — What it does
文本、音频、视频统一到统一坐标,输出 0-1 一致性评分与冲突分层。当前实时接通 财联社 · 公众号 · 抖音 · YouTube 四大渠道,雪球 / 微博 / 小红书等持续扩展中。 Text, audio, and video are normalized into a shared factor space. Output is a 0-1 consistency score plus conflict tiers. 4 channels are live now (CLS, WeChat MP, Douyin, YouTube); Xueqiu, Weibo, Xiaohongshu are being onboarded.
不以对话替代判断,而是把一句问题转成可调度研究任务。语音和文本都可输入,返回 JSON(source_map、signal、reason、confidence)。 Not chit-chat. A question becomes a structured task. Both voice and text are accepted, returning JSON (source_map, signal, reason, confidence).
Alpha · Research · Risk 三 Agent 通过 asyncio.gather 并发跑同一份融合上下文,故障隔离 + 8 秒超时。 Alpha · Research · Risk run via asyncio.gather over a shared fusion context. Failure-isolated, 8 s per-agent timeout.
输出不止 BUY / SELL。系统保留板块 verdict、龙头来源、多空配对与置信度,并可形成 {"signal": "pair_trade", "long": "...", "short": "...", "reason": "..."} 结构化输出。 Not just BUY / SELL. The system keeps sector verdicts, leader provenance, pair trade tags, and confidence weights in a structured output like {"signal": "pair_trade", "long": "...", "short": "...", "reason": "..."}.
夜盘、休市、海外时段持续采集与监控,关键脉冲触发推送 — 不漏凌晨那一条。 Continuous collection through after-hours, holidays and overseas sessions. Every meaningful pulse pushes to you — never miss the 3 AM one.
03 — 谁在用03 — Who it's for
3 人/天扑在社交平台上扒数据,因子覆盖率不到 20%,新因子出得比同行慢一个季度。 3 person-days a week scraping social. Factor coverage under 20%. New factors a quarter behind peers.
每天 30 分钟到 2 小时盯盘看消息,被自媒体拉扯,事后才发现真正的关键信号在另一个平台。 30 min – 2 hr a day chasing headlines. Whipsawed by influencers. The real signal was on a platform you didn't check.
行业事件爆发,需要在 1 小时内把跨平台叙事整理成可交付材料,人工不够。 Sector event hits. You need 4 platforms synthesized into a publishable note in 60 min. Manual doesn't scale.
* 数据为内部基准与早期合作伙伴脱敏样本,非签约客户公开数据。我们在内测期间会逐步用授权后的真实样本替换。 * Numbers are internal benchmarks and de-identified pilot samples. We replace them with authorised real samples as the beta progresses.
04 — 演示04 — See it move
问题输入(支持语音/文本)→ 跨平台采集 → 一致性与冲突标签 → 板块 verdict / 龙头候选 → 多空配对与仓位信号。我们会展示可复盘 evidence chain、因子快照与样本归因。 Query input (voice or text) → cross-platform collection → consistency and conflict labeling → sector verdicts / leaders → pair-trade signal and sizing. We show an auditable evidence chain, factor snapshots, and attribution sample windows.
提前拿到演示Get the demo first05 — API 交付05 — API Delivery
把语音/文本输入先转为结构化因子,再按多空对得到回测归因,直接对接回测引擎。 First produce structured factor output, then request attribution on the pair for backtest-ready inputs.
factor_signals JSON
{"ok":false,"msg":"click to run"}
可直接复用来自上一步的 pair,或手动填写。 Can reuse pair from step 1 or edit manually.
attribution JSON
{"ok":false,"msg":"run attribution after signal"}
在同一视图对齐: Align in a single view:
05.1 — 接口交付05.1 — API delivery
每次返回都包含 ok/metadata/source_rows/attribution_windows 等可落地字段。以下为当前生产可用请求方式。
Each response includes operational fields such as ok, metadata, source_rows and attribution_windows. Use as-is for integration now.
请求:GET /api/factor-signals?query=...§or=...&mode=demo|prod|live
Request: GET /api/factor-signals?query=...§or=...&mode=demo|prod|live
{"query":"新能源 AI 供应链预期差","sector":"新能源"}
curl -X GET "https://alphaplus-landing.datapro.asia/api/factor-signals?query=%E6%96%B0%E8%83%BD%E8%83%BD§or=%E5%8D%8E%E7%AB%8B%E9%80%9A"
请求:GET /api/attribution?pair=A/B&windows=5,10,20,30&mode=demo|prod|live
Request: GET /api/attribution?pair=A/B&windows=5,10,20,30&mode=demo|prod|live
{ "pair": { "long": "BYD", "short": "TSLA" }, "attribution_windows": [...], "factor_snapshot": {"pair_signal_id":"..."} }
curl -X GET "https://alphaplus-landing.datapro.asia/api/attribution?pair=BYD/TSLA&windows=5,10,20,30"
请求:GET /api/closed-loop?query=...§or=...&windows=5,10,20,30&mode=demo|prod|live
Request: GET /api/closed-loop?query=...§or=...&windows=5,10,20,30&mode=demo|prod|live
返回:factor + attribution + decision 一次性闭环包,前端可直接用于归因复核与回测接入。
Response: one-shot bundle factor + attribution + decision for review and backtest integration.
{
"ok": true,
"decision": {
"pair": "AMD/00700",
"consistency_score": 0.71,
"conflict_level": "low",
"best_window_days": 10,
"best_sharpe": 2.181,
"ready_for_backtest": false
}
}
curl -X GET "https://alphaplus-landing.datapro.asia/api/closed-loop?query=%E6%96%B0%E8%83%BDAI§or=%E5%8D%8E%E7%AB%8B%E9%80%9A&windows=5,10,20,30"
请求:POST /api/mcp,支持 tool 或 MCP tools/call 标准体
Request: POST /api/mcp, supports tool payload or MCP tools/call body
curl -X POST https://alphaplus-landing.datapro.asia/api/mcp -H "Content-Type: application/json" -d "{\"tool\":\"alpha_pulse.generate_factors\",\"arguments\":{\"query\":\"新能源 AI\",\"sector\":\"AI 处理器\",\"mode\":\"demo\"}}"
curl -X POST https://alphaplus-landing.datapro.asia/api/mcp -H "Content-Type: application/json" -d "{\"name\":\"alpha_pulse.get_attribution\",\"arguments\":{\"pair\":\"BYD/TSLA\",\"windows\":[5,10,20,30]}}"
以下字段可直接映射到研究复核与风控告警系统,形成“单次查询即交付”流程: These fields map directly into research audit and risk-policy systems for single-call handoff.
factor.pair / factor.signal.long/short(买入池/做空池):用于入库交易标的与多空配对。decision.consistency_score:一致性评分,触发前置风控阈值。decision.conflict_level:冲突级别,映射到风险分层规则。attribution.attribution_windows[].{window_days, win_rate, sharpe, max_drawdown}:作为回测归因报表 KPI。factor.source_rows[]:证据链与证据来源,作为合规留痕。{
"alpha_signal_id": "sig_TSLA_BYD_20260506",
"pair": {"long":"TSLA","short":"BYD"},
"consistency_score": 0.71,
"conflict_level": "high",
"best_window_days": 10,
"best_sharpe": 2.18,
"best_win_rate": 0.71,
"attribution_windows": [
{"window_days":10,"win_rate":0.71,"sharpe":2.18,"max_drawdown":-0.084}
],
"source_rows_count": 28
}
curl -X POST https://alphaplus-landing.datapro.asia/api/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","method":"tools/call","params":{"name":"alpha_pulse.run_closed_loop","arguments":{"query":"AI","sector":"新能源汽车","windows":[5,10,20,30]}}}'
const factor = await fetch('/api/factor-signals?query=...' ).then((r) => r.json());
const pair = `${factor.pair?.long}/${factor.pair?.short}`;
const attribution = await fetch(`/api/attribution?pair=${encodeURIComponent(pair)}`).then((r) => r.json());
// 使用 attribution.attribution_windows 与 attribution.equity_curves 进入回测流水线
06.5 — 齐总汇报包06.5 — Investor brief pack
点击复制后可直接粘贴给齐总:展示闭环参数、决策字段和回测归因。默认输出支持 json/md/csv,便于复盘和风控联调。 Copy this and paste directly for investor review: show loop input, decision fields, and attribution. Output supports json/md/csv for audit and compliance handoff.
# 请先点击「导出闭环报告」生成最新结果
# Please run "Export closed-loop report" first to generate real data
06 — 价格06 — Pricing
¥0 /月/mo
¥99 /月/mo
¥299 /月/mo
机构 / 量化私募 / 券商可协商「基础服务费 + 超额收益分成」,先做三个月闭环验证。 Institutional / quant fund / sell-side pilots can co-design with base fee + performance share, validated over 3-month evidence loops.
07 — 加入内测07 — Join the beta
填写一分钟,先给你发送可落地的研究样本:一致性冲突评分、配对信号和可复盘字段示例。早期用户优先进入闭环反馈名单。 One minute to fill in. You’ll receive practical samples first: consistency/conflict scores, pair signals, and auditable workflow fields.