360 Secure SkillHub
360 安全技能中心

Numinous Forecast

该 Skill 通过 Python 脚本调用 Numinous 预测 API，支持 query 模式（自然语言问题）与 event 模式（标题、描述、截止时间、主题），返回 0 到 1 的校准概率，以及 miner、推理文本、版本号、事件标题等来源元数据。每次请求通过 x402 按次付费，需配置钱包私钥环境变量。

juandbalbi
juandbalbi
热度 65.42
最近更新 2026-05-11 17:26:00
分类：数据依赖：代码执行 / 第三方 API应用场景：研究 / 市场分析

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add numinous-forecast
Metadata
name
numinous-forecast
description
Get calibrated probabilities from Numinous (Bittensor Subnet 6) with metadata/provenance.
homepage
https://numinouslabs.io/
metadata
{"clawdbot":{"emoji":"🔮","homepage":"https://numinouslabs.io/","requires":{"bins":["python3"],"env":["NUMINOUS_X402_EVM_PRIVATE_KEY"]},"primaryEnv":"NUMINOUS_X402_EVM_PRIVATE_KEY","install":[{"id":"uv-evm","kind":"uv","module":"x402[httpx,evm]","label":"Install x402 (httpx + EVM) (uv)"},{"id":"uv-evm-svm","kind":"uv","module":"x402[httpx,evm,svm]","label":"Install x402 (httpx + EVM + SVM) (uv)"}]}}

Numinous Forecast

This skill calls the Numinous forecasting API and returns a calibrated probability (p \in [0,1]) plus metadata/provenance.

Note: requests are paid per-call via x402 (HTTP 402 → pay → retry). You’ll need a wallet key configured (see setup).

Setup

Install Python deps (recommended: uv).

EVM-only (recommended):

uv pip install "x402[httpx,evm]"

If you also want Solana payments:

uv pip install "x402[httpx,evm,svm]"

Set the buyer key (required):

  • NUMINOUS_X402_EVM_PRIVATE_KEY: EVM key (0x…) for Base / EVM payments

Optional (Solana payments):

  • NUMINOUS_X402_SVM_PRIVATE_KEY: Solana key (base58) for Solana payments

Optional:

  • NUMINOUS_X402_PREFER: auto (default) | evm | svm

Security note: these are private keys. Treat them like cash. Don’t paste them into chats/logs.

Predict (query mode)

python3 "{baseDir}/predict_query.py" "Will BTC be above $100k by 2026-12-31?"
python3 "{baseDir}/predict_query.py" "Will Team X win League Y in 2026?" --topics sports

Predict (event mode)

python3 "{baseDir}/predict_event.py" \
  --title "Will BTC be above $100k by 2026-12-31?" \
  --description "Resolve YES if BTC/USD spot price is strictly above $100,000 at 2026-12-31 23:59:59 UTC." \
  --cutoff "2026-12-31T23:59:59Z" \
  --topics general

Output

Both scripts print JSON with these top-level fields:

  • ok: true|false
  • prediction: probability (p \in [0, 1])
  • forecasted_at: ISO timestamp (UTC)
  • forecaster_name: forecaster identifier
  • metadata: provenance/debug context (see below)
  • parsed_fields: query mode only — structured event parsed from your query (see below)
  • error: null on success; otherwise an error string

metadata (common keys)

  • pool: aggregation pool name
  • miner_uid: miner UID chosen for this forecast
  • miner_hotkey: miner hotkey identifier
  • reasoning: model reasoning / explanation text
  • agent_name: which forecaster agent produced the result
  • version_id: model/version identifier
  • version_number: integer version
  • raw_prediction: original prediction value before any formatting
  • event_title: resolved event title
  • event_cutoff: resolved cutoff timestamp (UTC)

parsed_fields (query mode only)

  • title: resolved event title
  • description: resolution criteria / event description
  • cutoff: ISO cutoff (often ends with Z)
  • topics: list of topic tags