Content Signal Radar

Not a news digest. A decision system built on top of your feed.

Built from a fork of follow-builders, but redesigned from the ground up to answer a different set of questions.

中文文档

What it does

Runs daily via GitHub Actions. Pulls signals from multiple source types, scores them, filters noise, and produces a structured signal report — ready for stdout, Telegram, dashboard integration, or repo-based archiving. It helps you decide:

• What should I pay attention to today?
• What can I write about on X?
• What can become a Xiaohongshu post?
• What should influence my product or workflow decisions?
• What's my concrete next action?

Sources supported

Type	Method
X (Twitter)	RSS via nitter instances
Blogs / newsletters	RSS
Podcasts	RSS
即刻 (Jike)	API
公众号 (WeChat MP)	wewe-rss

Sources are split into config/default-sources.json (public, version-controlled) and config/custom-sources.json (private, git-ignored). Add your own accounts in the custom file.

How it works

fetch feeds
  → deduplicate (seen-signals with daily TTL reset, Asia/Shanghai timezone)
  → score signals (focusTopics × contentGoals × scoring config)
  → prepare digest (platform sections: X drafts, Xiaohongshu angles, product signals)
  → deliver via stdout / Telegram
  → write dashboard-signals.json (for dashboard integration)
  → commit feed state back to repo

Key scripts:

• scripts/generate-feed.js — fetches and deduplicates all sources
• scripts/prepare-digest.js — scores, sections, and formats the report
• scripts/deliver.js — pushes to Telegram

Configuration

Copy config/maple.config.example.json → config/maple.config.json (git-ignored).

Key fields:

{
  "focusTopics": ["AI agents", "creator economy", "personal brand"],
  "contentGoals": ["X post drafts", "Xiaohongshu topics", "product signals"],
  "outputSections": ["signals", "x-drafts", "xiaohongshu", "product"],
  "scoring": {
    "minScore": 3,
    "boostKeywords": ["agent", "workflow", "creator"]
  }
}

Full schema: config/config-schema.json

Setup

1. Fork this repo
2. Add repository secrets used by feed generation:
   • X_BEARER_TOKEN
   • SUPADATA_API_KEY
3. For local report preparation / delivery, copy config/maple.config.example.json to ~/.content-signal-radar/config.json
4. Add private sources to ~/.content-signal-radar/custom-sources.json (see config/custom-sources.example.json)
5. GitHub Actions runs automatically once daily via .github/workflows/generate-feed.yml

Optional Telegram delivery uses TELEGRAM_BOT_TOKEN in ~/.content-signal-radar/.env plus delivery.chatId in your local config.

See examples/sample-digest.md for a sample output.

Output

Output	Description
stdout / Telegram	Structured signal report, depending on delivery config
dashboard-signals.json	Structured signal data for external integrations
output/	Historical reports committed to the repo when enabled

Credits

Original project: zarazhangrui/follow-builders This fork: signal extraction over generic summarization, tuned for content + product decisions.