Skip to main content

browse_read

The browse_read tool extracts and reads text content from papers and other content sources. If the content has not been downloaded yet, it will automatically download it before extracting text.

Basic Usage​

browse_read(searcher="arxiv", paper_id="2303.08774")

Parameters​

ParameterTypeRequiredDefaultDescription
searcherstringYes-Source platform
paper_idstringYes-Content identifier (1-200 characters)
pageintegerNo-Read specific page (1-indexed)
start_pageintegerNo-Page range start (1-indexed)
end_pageintegerNo-Page range end (1-indexed)

Pagination​

The browse_read tool supports reading specific pages or page ranges from PDF documents. This is useful for:

  • Reading only specific sections without loading the entire document
  • Efficiently browsing long papers
  • Reducing context length for AI assistants

Pagination Parameters​

ParameterDescriptionExample
pageRead a single specific pagepage=3 returns only page 3
start_pagePage range start (inclusive)start_page=1 starts from page 1
end_pagePage range end (inclusive)end_page=5 ends at page 5

Pagination Behavior​

ParametersResult
NoneReturns all pages
page=3Returns only page 3
start_page=1, end_page=5Returns pages 1-5
start_page=10From page 10 to end
end_page=5Returns pages 1-5

Pagination Examples​

# Read only abstract (usually page 1)
browse_read(searcher="arxiv", paper_id="2303.08774", page=1)

# Read introduction (pages 1-3)
browse_read(searcher="arxiv", paper_id="2303.08774", start_page=1, end_page=3)

# Start reading from methods section (assuming it starts at page 5)
browse_read(searcher="arxiv", paper_id="2303.08774", start_page=5)

# Read up to conclusion (first 10 pages)
browse_read(searcher="arxiv", paper_id="2303.08774", end_page=10)

Pagination Response Format​

When using pagination, the response includes page markers:

--- Page 1 ---
Title: GPT-4 Technical Report

Abstract
We report the development of GPT-4, a large-scale, multimodal
model which can accept image and text inputs...

--- Page 2 ---
1 Introduction
This technical report presents GPT-4, a large multimodal model
capable of processing image and text inputs...

Paper ID Formats​

Each platform uses different identifier formats. See browse_download for complete format details.

SearcherExample
arxiv2303.08774
pubmed32790614
pmcPMC7419405
biorxiv10.1101/2020.01.01.123456
medrxiv10.1101/2020.01.01.123456
iacr2009/101
crossref10.1038/s41586-020-2649-2
semanticDOI:10.18653/v1/N18-3011
core123456789

Reading Examples​

Read from Different Data Sources​

# Read from arXiv
browse_read(searcher="arxiv", paper_id="2106.12345")

# Read from PubMed
browse_read(searcher="pubmed", paper_id="32790614")

# Read from PubMed Central
browse_read(searcher="pmc", paper_id="PMC7419405")

# Read from bioRxiv
browse_read(searcher="biorxiv", paper_id="10.1101/2020.01.01.123456")

# Read from medRxiv
browse_read(searcher="medrxiv", paper_id="10.1101/2020.01.01.123456")

# Read from IACR
browse_read(searcher="iacr", paper_id="2009/101")

# Read from Semantic Scholar
browse_read(searcher="semantic", paper_id="DOI:10.18653/v1/N18-3011")

# Read from CrossRef
browse_read(searcher="crossref", paper_id="10.1038/s41586-020-2649-2")

# Read from CORE
browse_read(searcher="core", paper_id="123456789")

Read from Plugin Data Sources​

If the social media plugin is installed:

# Read from GitHub
browse_read(searcher="github", paper_id="owner/repo")

# Read from Twitter
browse_read(searcher="twitter", paper_id="1234567890")

# Read from Zhihu
browse_read(searcher="zhihu", paper_id="123456789")

How It Works​

  1. Check local cache: The tool first checks if the content has been downloaded
  2. Download if needed: If not available locally, automatically downloads the content
  3. Extract text: Uses appropriate parser (PDF, HTML, etc.) to extract text
  4. Apply pagination: If pagination parameters are set, extracts only the requested pages
  5. Return content: Returns the extracted text string

Response Format​

The tool returns extracted text content:

Title: GPT-4 Technical Report

Abstract
We report the development of GPT-4, a large-scale, multimodal
model which can accept image and text inputs and produce text
outputs. While less capable than humans in many real-world
scenarios, GPT-4 exhibits human-level performance on various
professional and academic benchmarks...

1 Introduction
This technical report presents GPT-4, a large multimodal model
capable of processing image and text inputs and producing text
outputs...

[Full paper text continues...]

Input Validation​

  • searcher: Must be one of the enabled data sources
  • paper_id: Must be 1-200 characters, cannot be empty or whitespace only
  • page: Must be a positive integer (1 or greater)
  • start_page: Must be a positive integer (1 or greater)
  • end_page: Must be a positive integer, greater than or equal to start_page

Error Handling​

Common errors and their meanings:

ErrorCauseSolution
Searcher unavailableData source not enabledEnable the data source in configuration
Paper ID cannot be emptyEmpty or whitespace only IDProvide a valid paper ID
Paper not foundInvalid paper IDVerify paper ID format
Error converting paper to textPDF parsing failedTry re-downloading or use another data source
Invalid page numberPage number out of rangeUse valid page numbers

Tips​

:::tip Workflow For best results, first use browse_search to search for papers, then use the returned paper ID with browse_read to extract content. :::

:::tip Pagination for Long Papers For long papers, use pagination to read specific sections:

  • page=1 to get abstract

  • start_page=1, end_page=3 to get introduction

  • Only read full paper when needed :::

  • The tool automatically downloads papers, so you don't need to call browse_download first

  • Downloaded papers are cached for faster subsequent reads

  • Text extraction quality depends on PDF structure (some scanned PDFs may have poor extraction results)

  • Pagination only works for PDF content; other content types return full text

Use Cases​

Research Summary​

Ask your AI assistant:

"Read page 1 of paper 2303.08774 from arXiv and summarize the abstract"

Literature Review​

After searching:

"Search for papers about transformer architecture on arXiv, then read pages 1-5 of the top result"

Citation Extraction​

"Read the last 3 pages of this paper to find the references section"

Progressive Reading​

"Read pages 1-5 first, if more details are needed, then read pages 6-10"

Next Steps​