📰 Hacker News Agent

An intelligent agent built with nanoagent that curates personalized Hacker News digests based on your interests.

✨ Features

15+ LLM Providers - OpenAI, Anthropic, DeepSeek, Groq, Ollama, and many more
Extremely Affordable - Use DeepSeek at $0.14/1M tokens (400x cheaper than GPT-4!)
Free Options - OpenRouter free tier or 100% free local Ollama
Smart Story Fetching - Retrieves top, new, best, or trending stories from Hacker News
Intelligent Filtering - Filters by score, comments, keywords, and time range
AI-Powered Analysis - Analyzes stories for relevance and importance
Customizable Preferences - Define your tech interests and filtering criteria
Multiple Output Formats - Markdown, JSON, or plain text reports
Automated Daily Digests - Can be scheduled to run daily
Privacy-Focused - Option to run 100% locally with Ollama

🚀 Quick Start

Prerequisites

Bun >= 1.2.0 (or Node.js >= 18)
An LLM API key (see LLM Providers below) OR local Ollama installation

Installation

# Install dependencies
bun install
# or with npm
npm install

# Set up environment variables
cp .env.example .env
# Edit .env and add your API key

Run the Agent

# Run with default (DeepSeek - cheapest)
bun run dev

# Run with specific examples
bun run examples/daily-hn-cheap.ts  # Affordable models
bun run examples/daily-hn-ollama.ts # Local/free models
bun run examples/compare-models.ts  # Compare different models

# With npm
npm run dev

📖 Usage

Basic Example

import { runHNAgent, extractReport } from "./src/index.ts";

const result = await runHNAgent({
  model: {
    provider: "openai",
    model: "gpt-4",
    apiKey: process.env.OPENAI_API_KEY,
  },
  preferences: {
    interests: ["AI", "machine learning", "web development"],
    min_score: 100,
    max_stories: 10,
    report_format: "markdown",
  },
});

const report = extractReport(result);
console.log(report);

Configuration Options

interface HNAgentConfig {
  model: any; // LLM model configuration
  preferences?: {
    interests: string[]; // Topics you're interested in
    min_score: number; // Minimum story score (default: 50)
    min_comments: number; // Minimum comments (default: 10)
    exclude_keywords: string[]; // Keywords to exclude
    max_stories: number; // Max stories in report (default: 10)
    report_format: "markdown" | "json" | "text";
    time_range_hours: number; // Only recent stories (default: 24)
  };
  storyType?: "top" | "new" | "best" | "ask" | "show";
  maxIterations?: number; // Max agent steps (default: 20)
}

Using Different Models

import { runHNAgent } from "./src/index.ts";
import { Recommended, DeepSeek, Groq, Ollama } from "./src/config/models.ts";

// 🏆 Recommended: DeepSeek (Best value)
const result = await runHNAgent({ model: Recommended.CHEAPEST });

// 🆓 Free: OpenRouter or Ollama
const result = await runHNAgent({ model: Recommended.FREE });

// ⚡ Fastest: Groq
const result = await runHNAgent({ model: Recommended.FASTEST });

// 🏠 Private: Local Ollama
const result = await runHNAgent({ model: Ollama.LLAMA_3_2 });

// 💎 Premium: OpenAI/Anthropic
const result = await runHNAgent({ model: OpenAI.GPT4 });

See LLM Providers section below for complete list and pricing.

🛠️ Project Structure

nanoagent/
├── src/
│   ├── agents/
│   │   └── hackernews-agent.ts    # Main agent orchestration
│   ├── tools/
│   │   ├── fetch-hn-stories.ts    # Fetch stories from HN API
│   │   ├── filter-stories.ts      # Filter and rank stories
│   │   ├── analyze-story.ts       # AI story analysis
│   │   └── generate-report.ts     # Report generation
│   ├── utils/
│   │   ├── hn-api.ts              # HN API client
│   │   ├── scoring.ts             # Scoring algorithms
│   │   └── formatting.ts          # Output formatting
│   ├── types/
│   │   └── hackernews.ts          # TypeScript types
│   └── index.ts                   # Main exports
├── examples/
│   └── daily-hn.ts                # Example usage
└── package.json

🔧 Development

# Run tests
bun test

# Lint code
bun run lint

# Format code
bun run format

# Build
bun run build

📋 Agent Workflow

The agent follows this workflow:

Fetch Stories - Retrieves stories from HN API using fetch_hn_stories tool
Filter Stories - Applies user criteria using filter_stories tool
Analyze Stories - Analyzes top stories using analyze_story tool
Generate Report - Creates formatted report using generate_report tool

🎯 Example Output

# 📰 Your Daily Hacker News Digest

**Date:** Friday, November 7, 2025
**Stories Found:** 10

---

### 1. New Claude 4 Released with Improved Reasoning

**Score:** 847 | **Comments:** 312 | **Posted:** Fri Nov 07 2025 08:23:15
**Topics:** AI, machine learning
**Relevance:** 95%

**Why it matters:** Highly popular with significant community engagement;
active discussion with diverse perspectives

**Summary:** Story about AI, machine learning with 847 points and 312 comments.

**Links:** [Article](https://anthropic.com/...) | [HN Discussion](https://news.ycombinator.com/item?id=...)

---

🔄 Scheduling Daily Digests

You can schedule the agent to run daily using cron:

# Add to crontab (runs daily at 8 AM)
0 8 * * * cd /path/to/nanoagent && bun run examples/daily-hn.ts >> ~/hn-digest.log 2>&1

Or use a task scheduler on Windows.

🤖 LLM Providers Support

We support 15+ LLM providers including many affordable and free options!

💰 Cost Comparison

Provider	Model	Cost per 1M tokens	Speed	Quality
DeepSeek ⭐	deepseek-chat	$0.14	Fast	Excellent
SiliconFlow	Qwen2.5-7B	$0.20	Fast	Very Good
OpenRouter	Free models	FREE	Medium	Good
Groq	Llama 3.1 8B	FREE tier	Super Fast	Very Good
Ollama 🏠	Any model	FREE	Varies	Good-Excellent
Together AI	Llama 3.1 70B	$0.88	Fast	Excellent
Moonshot	moonshot-v1-8k	$1.20	Medium	Very Good
GLM	glm-4	$1.50	Medium	Very Good
OpenAI	GPT-3.5 Turbo	$1.50	Fast	Excellent
OpenAI	GPT-4	$60.00	Medium	Excellent
Anthropic	Claude 3 Sonnet	$15.00	Fast	Excellent

⭐ = Recommended for best value 🏠 = Runs locally (100% free & private)

📝 Setup Instructions by Provider

🏆 DeepSeek (Recommended - Cheapest!)

# Sign up at https://platform.deepseek.com/
export DEEPSEEK_API_KEY="your-key"

import { DeepSeek } from "./src/config/models.ts";
const result = await runHNAgent({ model: DeepSeek.CHAT });

Pricing: ~$0.14 per 1M tokens (400x cheaper than GPT-4!) Quality: Excellent, comparable to GPT-3.5 Turbo Best for: Daily use, high-volume tasks

🆓 OpenRouter (Free Models Available)

# Sign up at https://openrouter.ai/
export OPENROUTER_API_KEY="your-key"

import { OpenRouter } from "./src/config/models.ts";

// Free models
const result = await runHNAgent({ model: OpenRouter.LLAMA_3_1_8B_FREE });

// Or paid models (still cheap)
const result = await runHNAgent({ model: OpenRouter.DEEPSEEK_CHAT });

Pricing: FREE tier available, paid models from $0.20/1M Benefits: Access 100+ models through one API Best for: Trying different models, free tier users

⚡ Groq (Super Fast!)

# Sign up at https://groq.com/
export GROQ_API_KEY="your-key"

import { Groq } from "./src/config/models.ts";
const result = await runHNAgent({ model: Groq.LLAMA_3_1_70B });

Pricing: FREE tier available, then paid Speed: Up to 750 tokens/second (fastest!) Best for: Real-time applications, quick responses

🏠 Ollama (100% Free & Private)

# Install from https://ollama.ai/
curl -fsSL https://ollama.ai/install.sh | sh

# Start Ollama
ollama serve

# Pull a model
ollama pull llama3.2

import { Ollama } from "./src/config/models.ts";
const result = await runHNAgent({ model: Ollama.LLAMA_3_2 });

Pricing: FREE (runs on your computer) Privacy: 100% private, no data sent to cloud Best for: Privacy, offline use, unlimited usage

Available models:

llama3.2 - Good balance of speed/quality
qwen2.5 - Excellent for general use
deepseek-coder-v2 - Best for technical content
mistral - Fast and efficient
gemma2 - Google's model

🇨🇳 Chinese Providers (Affordable)

Moonshot AI:

export MOONSHOT_API_KEY="your-key"

Long context support (up to 200K tokens)
Sign up: https://platform.moonshot.cn/

SiliconFlow:

export SILICONFLOW_API_KEY="your-key"

Hosts many open source models
Sign up: https://siliconflow.cn/

GLM (Zhipu AI):

export GLM_API_KEY="your-key"

ChatGLM models
Sign up: https://open.bigmodel.cn/

💎 Premium Providers

OpenAI:

export OPENAI_API_KEY="sk-..."

Anthropic Claude:

export ANTHROPIC_API_KEY="sk-ant-..."

🎯 Recommended Configurations

import { Recommended } from "./src/config/models.ts";

// Best for cost-effectiveness
const model = Recommended.CHEAPEST;  // DeepSeek

// Best free option
const model = Recommended.FREE;  // OpenRouter free tier

// Best for privacy
const model = Recommended.PRIVATE;  // Ollama local

// Best for speed
const model = Recommended.FASTEST;  // Groq

// Best for coding/tech content
const model = Recommended.CODING;  // DeepSeek Coder

// Best for Chinese language
const model = Recommended.CHINESE;  // Moonshot

📊 Benchmark Your Models

Compare different models with the benchmark script:

bun run examples/compare-models.ts

This will test multiple models and show:

Response time
Quality comparison
Cost estimates

🔄 Scheduling Daily Digests

Using Cron (Linux/Mac)

# Edit crontab
crontab -e

# Add this line (runs daily at 8 AM)
0 8 * * * cd /path/to/nanoagent && bun run examples/daily-hn-cheap.ts >> ~/hn-digest.log 2>&1

Using Windows Task Scheduler

Open Task Scheduler
Create Basic Task
Set trigger: Daily at 8:00 AM
Action: Start a program
Program: bun
Arguments: run examples/daily-hn-cheap.ts
Start in: /path/to/nanoagent

Using systemd Timer (Linux)

# Create ~/.config/systemd/user/hn-digest.service
[Unit]
Description=Daily HN Digest

[Service]
Type=oneshot
WorkingDirectory=/path/to/nanoagent
ExecStart=/usr/bin/bun run examples/daily-hn-cheap.ts

# Create ~/.config/systemd/user/hn-digest.timer
[Unit]
Description=Daily HN Digest Timer

[Timer]
OnCalendar=daily
OnCalendar=08:00
Persistent=true

[Install]
WantedBy=timers.target

# Enable and start
systemctl --user enable hn-digest.timer
systemctl --user start hn-digest.timer

🤝 Contributing

Contributions are welcome! Please feel free to submit issues or pull requests.

📄 License

MIT

🙏 Acknowledgments

Built with nanoagent
Uses the official Hacker News API
Supports 15+ LLM providers for maximum flexibility

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
examples		examples
src		src
.env.example		.env.example
.gitignore		.gitignore
PROVIDERS.md		PROVIDERS.md
QUICK_START.md		QUICK_START.md
README.md		README.md
WINDOWS.md		WINDOWS.md
biome.json		biome.json
config.example.json		config.example.json
package-lock.json		package-lock.json
package.json		package.json
tsconfig.json		tsconfig.json

ChopperLin/nanoagent

Folders and files

Latest commit

History

Repository files navigation

📰 Hacker News Agent

✨ Features

🚀 Quick Start

Prerequisites

Installation

Run the Agent

📖 Usage

Basic Example

Configuration Options

Using Different Models

🛠️ Project Structure

🔧 Development

📋 Agent Workflow

🎯 Example Output

🔄 Scheduling Daily Digests

🤖 LLM Providers Support

💰 Cost Comparison

📝 Setup Instructions by Provider

🏆 DeepSeek (Recommended - Cheapest!)

🆓 OpenRouter (Free Models Available)

⚡ Groq (Super Fast!)

🏠 Ollama (100% Free & Private)

🇨🇳 Chinese Providers (Affordable)

💎 Premium Providers

🎯 Recommended Configurations

📊 Benchmark Your Models

🔄 Scheduling Daily Digests

Using Cron (Linux/Mac)

Using Windows Task Scheduler

Using systemd Timer (Linux)

🤝 Contributing

📄 License

🙏 Acknowledgments

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages