How to Build a Research Assistant for GitHub, arXiv, and Hacker News

Some questions need more than one source type. A research assistant should look at code, papers, and discussion threads before it answers.

That is exactly where a multi-engine search API helps.

The use case

This pattern is useful when the question is:

• "What are people building around this topic?"
• "Which repo has the implementation details?"
• "What papers are current on this subject?"
• "Is the community discussing this already?"

GitHub, arXiv, and Hacker News are a strong combination because they cover implementation, research, and commentary.

Minimal Prismfy query

curl -X POST https://api.prismfy.io/v1/search \
  -H "Authorization: Bearer ss_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "agentic search RAG",
    "engines": ["github", "arxiv", "hackernews"],
    "timeRange": "month",
    "language": "en"
  }'

Turning results into a research packet

import requests

def research_packet(topic: str) -> dict:
    response = requests.post(
        "https://api.prismfy.io/v1/search",
        headers={"Authorization": "Bearer ss_live_YOUR_KEY"},
        json={
            "query": topic,
            "engines": ["github", "arxiv", "hackernews"],
            "timeRange": "month",
            "language": "en",
        },
        timeout=30,
    )
    response.raise_for_status()
    data = response.json()

    return {
        "topic": topic,
        "top_results": [
            {
                "title": r["title"],
                "url": r["url"],
                "snippet": r.get("content", "")[:240],
            }
            for r in data.get("results", [])[:5]
        ],
    }

packet = research_packet("agentic search RAG")
print(packet)

How to use it

A good research assistant should not answer immediately. It should:

1. Gather code examples from GitHub.
2. Pull recent papers from arXiv.
3. Check what the community is saying on Hacker News.
4. Summarize the evidence and cite the sources.

That is enough for a useful internal research workflow.

Common mistakes

• Relying on a single source type for a mixed research question.
• Returning a giant dump of links instead of a short evidence pack.
• Ignoring freshness and using only old papers or stale repos.
• Writing a "research assistant" article that never shows the actual search call.

If you need a research assistant that can blend code, papers, and discussion, start with Prismfy docs and test a multi-engine POST /v1/search query.