Skip to content

Quick Start

Get up and running with ScoutML in 5 minutes!

Prerequisites

Before starting, ensure you have:

Choose Your Interface

ScoutML provides two ways to interact with the API:

Let's start by searching for papers about transformers:

scoutml search "transformer models" --limit 5

This returns the top 5 papers matching your query, displayed in a beautiful table format.

Your First Search

Let's start by searching for papers about transformers:

import scoutml

# Search for papers
results = scoutml.search("transformer models", limit=5)

# Display results
for paper in results['papers']:
    print(f"• {paper['title']} ({paper['year']})")
    print(f"  Citations: {paper['citations']}")
    print(f"  ArXiv ID: {paper['arxiv_id']}\n")

Understanding the Output

The search results show:

  • Title: Paper title with clickable ArXiv link
  • Authors: First author et al.
  • Year: Publication year
  • Citations: Current citation count
  • Score: Relevance score (0-100)

Add filters to narrow down results:

# Papers from 2022 onwards with 50+ citations
scoutml search "vision transformer" \
  --year-min 2022 \
  --min-citations 50 \
  --limit 10

# Papers from 2022 onwards with 50+ citations
results = scoutml.search(
    "vision transformer",
    year_min=2022,
    min_citations=50,
    limit=10
)

Getting Paper Details

Found an interesting paper? Get more details:

# Using the ArXiv ID from search results
scoutml paper 2010.11929

# Using the ArXiv ID from search results
paper = scoutml.get_paper("2010.11929")

print(f"Title: {paper['paper']['title']}")
print(f"Authors: {', '.join(paper['paper']['authors'][:3])}...")
print(f"Abstract: {paper['paper']['abstract'][:200]}...")

This shows: - Full abstract - All authors - Publication details - Citation metrics - Related papers (use --similar flag)

Comparing Papers

Compare multiple papers side-by-side:

scoutml compare 1810.04805 2005.14165 1910.10683

comparison = scoutml.compare_papers("1810.04805", "2005.14165", "1910.10683")

print("Key Contributions:")
for paper_id, contrib in comparison['analysis']['contributions'].items():
    print(f"• {paper_id}: {contrib}")

The AI-powered comparison highlights: - Key contributions - Methodological differences - Performance comparisons - Strengths and limitations

Getting Implementation Help

Want to implement a paper? Our AI agent can help:

scoutml agent implement 2010.11929 --framework pytorch

guide = scoutml.get_implementation_guide("2010.11929", framework="pytorch")

print(f"Overview: {guide['implementation']['overview']}")
print("\nKey Components:")
for component in guide['implementation']['key_components']:
    print(f"• {component}")

This generates: - Step-by-step implementation guide - Code structure recommendations - Key components to implement - Common pitfalls to avoid

Finding Similar Papers

Discover related work:

scoutml similar --paper-id 1810.04805 --limit 5

similar = scoutml.find_similar_papers(paper_id="1810.04805", limit=5)

for paper in similar['papers']:
    print(f"• {paper['title']}")
    print(f"  Similarity: {paper['similarity']:.2%}")

Complete Workflow Example

Here's a complete research workflow:

# 1. Search for papers on a topic
scoutml search "self-supervised learning vision" --year-min 2021 --limit 10

# 2. Get details on an interesting paper
scoutml paper 2104.14294

# 3. Find similar papers
scoutml similar --paper-id 2104.14294 --limit 5

# 4. Compare top candidates
scoutml compare 2104.14294 2103.03230 2102.05918

# 5. Get implementation guide for the best one
scoutml agent implement 2104.14294 --framework pytorch

import scoutml

# 1. Search for papers on a topic
results = scoutml.search("self-supervised learning vision", year_min=2021, limit=10)
print(f"Found {len(results['papers'])} papers")

# 2. Get details on the most cited paper
top_paper = max(results['papers'], key=lambda p: p['citations'])
paper_id = top_paper['arxiv_id']
details = scoutml.get_paper(paper_id)

# 3. Find similar papers
similar = scoutml.find_similar_papers(paper_id=paper_id, limit=5)

# 4. Compare top candidates
candidate_ids = [p['arxiv_id'] for p in similar['papers'][:3]]
comparison = scoutml.compare_papers(*candidate_ids)

# 5. Get implementation guide for the best one
guide = scoutml.get_implementation_guide(paper_id, framework="pytorch")

Exporting Results

Save results for later analysis:

# Export search results to JSON
scoutml search "BERT" --output json --export bert_papers.json

# Export comparison to Markdown
scoutml compare 1810.04805 1906.08237 --output markdown --export comparison.md

What's Next?

Now that you know the basics:

Getting Help

If you run into issues: