Meta Agent Tools
PT

Back to the catalog

signal-hunter

Daily AI-powered trend hunter that aggregates content from 15+ sources — HackerNews, GitHub Trending, Reddit, YouTube (with transcripts), X/

Open source Repository Open in the app JSON README (API)

About

Daily AI-powered trend hunter that aggregates content from 15+ sources — HackerNews, GitHub Trending, Reddit, YouTube (with transcripts), X/Twitter, Product Hunt, HuggingFace Papers, Gmail newsletters, and app stores. Detects cross-source trending signals by finding topics appearing in 3+ sources simultaneously, rates them by strength, and generates structured Obsidian notes with business ideas (apps, SaaS, BaaS, gaming). Includes daily caching with historical data preservation to reduce API costs. Ships with 5 skills: /digest, /add-source, /test-source, /list-sources, and /setup.

Details

Kind
Plugins
Topic
Social & content
Publisher
berkkorkmaz
Origin
marketplace
Category
ferramentas
Stars
6
Last push
2026-04-07T10:57:16Z
Repository state
ativo
Language
Python
Added
2026-08-30 01:48:58
Updated
2026-08-30 01:48:58
Origin id
berkkorkmaz/signal-hunter/signal-hunter

README

<p align="center">
  <img src="assets/logo.png" alt="Signal Hunter" width="600" />
</p>

<p align="center">
  <strong>AI-powered daily trend hunter</strong><br>
  Aggregates 20+ sources, scores cross-source signals, generates Obsidian notes with business ideas.
</p>

---

## What it does

Every day, it collects from:

| Source | Method | What you get |
|--------|--------|-------------|
| Hacker News | WebFetch | Top stories with points and comments |
| Product Hunt | WebFetch | Top products with taglines and upvotes |
| GitHub Trending | WebFetch | Trending repos with stars and descriptions |
| HuggingFace | WebFetch | Trending papers with likes |
| smol.ai | WebFetch | AI news summaries |
| Superhuman AI | WebFetch | Curated AI newsletter content |
| Ben's Bites | WebFetch | AI startup and product coverage |
| TechCrunch AI | WebFetch | Enterprise AI news and funding |
| Futurepedia Innovations | WebFetch | Top 100 tech company AI developments |
| Reddit | Python API | Top 10 posts/day per subreddit (incl. LocalLLaMA) |
| YouTube | RSS + transcripts | New videos with summarized transcripts |
| X/Twitter | X API v2 | Latest tweets from accounts you follow |
| Gmail | Gmail MCP | Auto-discovered newsletter content |
| Futurepedia Tools | API Endpoint | New AI tool launches (structured JSON) |
| App stores | Playwright | Trending apps from AppMagic, AppRaven |

Then it:
1. **Scores** every item (0-100 normalized across sources)
2. **Detects cross-source signals** with 1.5x multiplier for topics in 2+ sources
3. **Tracks newsletter mentions** with +15 bonus (human-curated = strong signal)
4. **Tracks velocity** across 7 days (accelerating, new, fading, steady)
5. **Detects app store gaps** for trending topics with no existing apps
6. **Summarizes YouTube transcripts** into 5 bullet points + key quote
7. **Generates up to 3 business ideas** per day (app, SaaS, BaaS, gaming, dev-tools)
8. **Writes markdown notes** to any folder (works great with Obsidian, but not required)
9. **Sends email digests** via Gmail SMTP to you or a Google Group
10. **Caches everything** with historical data preserved per day

## Install

```bash
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
.venv/bin/playwright install chromium

cp config.yaml.example config.yaml
# edit config.yaml with your sources and output path

# optional: X/Twitter API and Gmail SMTP
echo "X_BEARER_TOKEN=your_token" > .env
echo "GMAIL_ADDRESS=you@gmail.com" >> .env
echo "GMAIL_APP_PASSWORD=xxxx xxxx xxxx xxxx" >> .env
```

Or run `/setup` in Claude Code — it also offers to install `/digest` as a global plugin so you can run it from the Claude desktop app or any directory.

## Usage

### As a Claude Code plugin

```
/digest              # run the full daily pipeline
/weekly              # generate weekly trend report
/add-source <url>    # add a new source (auto-categorized)
/test-source reddit  # test a category
/list-sources        # show all configured sources
```

### As a script

```bash
.venv/bin/python -m src.collector              # run all scrapers
.venv/bin/python -m src.scoring                # velocity + scores
.venv/bin/python -m src.scoring --weekly       # weekly aggregate
.venv/bin/python -m src.collector --test reddit # test one category
.venv/bin/python -m src.collector --force       # ignore daily cache
```

## Configuration

Copy `config.yaml.example` to `config.yaml` and customize:

```yaml
sources:
  reddit:
    - subreddit: singularity
    - subreddit: AINewsMinute

  youtube:
    - channel: "@AIDailyBrief"
      channel_id: UCKelCK4ZaO6HeEI1KQjqzWA

  twitter:
    - handle: nlw
    - handle: edsim

scoring:
  weights:
    hackernews: { divisor: 10 }
    reddit: { divisor: 5 }
    gmail: { default: 70 }
  cross_source_multiplier: 1.5
  newsletter_mention_bonus: 15

email:
  enabled: true
  recipients:
    - your-group@googlegroups.com

# Output directory — any folder works, Obsidian is optional
obsidian:
  vault_path: ~/Documents/signal-hunter-output
  daily_folder: Daily
  ideas_folder: Ideas
  weekly_folder: Weekly
```

### X/Twitter API (optional)

1. Go to [developer.x.com](https://developer.x.com)
2. Create a project, generate a Bearer Token
3. Add to `.env`: `X_BEARER_TOKEN=your_token`

### Gmail SMTP (optional)

1. Enable 2FA on your Google account
2. Generate an app password at [myaccount.google.com/apppasswords](https://myaccount.google.com/apppasswords)
3. Add to `.env`: `GMAIL_ADDRESS` and `GMAIL_APP_PASSWORD`

### Gmail MCP (optional)

For newsletter auto-discovery, enable the Claude Gmail plugin in Claude Code settings.

## Output

Markdown files written to your configured `vault_path` (any folder — Obsidian is optional):

```
output/
  Daily/
    2026-03-25.md                          # daily digest with scored trends
  Ideas/
    idea-2026-03-25-ai-security.md         # up to 3 ideas per day
    idea-2026-03-25-agent-monitoring.md
    idea-2026-03-25-task-manager.md
  Weekly/
    2026-W13.md                            # weekly rollup
```

Email digests are also sent to configured recipients with the same content.

## How trend detection works

The system catches emerging signals early by combining multiple detection methods:

1. **Score normalization** (0-100) across all source types
2. **Tiered cross-source multiplier** (1.3x for 2 sources, 1.5x for 3+) to reward true convergence
3. **Newsletter mention bonus** (+15) because human-curated newsletters are strong signal amplifiers
4. **Velocity tracking** against 7 days of cached data
5. **Cross-day deduplication** — each topic is marked `fresh`, `deepened` (new angle/sources), or `repeat` (same story, skipped). Only genuinely new or evolved stories surface in the digest.
6. **App store gap detection** to find market opportunities
7. **Weekly rollups** categorizing persistent trends, flash signals, and rising topics

## License

MIT

More