Kagi API Client for R • kagiPro

R client for the Kagi API.

Overview

kagiPro provides a lightweight R interface to the Kagi API, including:

Search API — perform web searches with advanced operators
Enrich API — get higher-signal results from specialized indices
Universal Summarizer — summarize text or URLs in one call
FastGPT API — ask grounded LLM questions with optional web context

The package follows the rOpenSci style for API clients:

S3 classes for connections, requests, and results
Extractor helpers to work with results as tibbles or text
Secure API key handling with keyring
Replay metadata for query-by-name reruns in project workflows

Installation

# Install the development version from GitHub
# install.packages("remotes")
remotes::install_github("rkrug/kagiPro")

Authentication

You need a Kagi account with API access (paid plan).
Store your API key securely in your system keychain:

# Run once to save your key in the keychain
keyring::key_set("API_kagi")

The package will resolve the key at request time with:

conn <- kagi_connection(
  api_key  = function() keyring::key_get("API_kagi")
)

Example

library(kagiPro)

# Build a query
q <- query_search(
  query    = 'biodiversity "annual report"',
  filetype = "pdf",
  site     = "example.com",
  expand   = FALSE
)

# Execute request and write JSON output
conn <- kagi_connection(api_key = function() keyring::key_get("API_kagi"))
out <- tempfile("kagiPro-search-")
dir.create(out, recursive = TRUE, showWarnings = FALSE)

kagi_request(
  connection = conn,
  query = q,
  limit = 3,
  output = out,
  overwrite = TRUE
)

Documentation

A detailed Quickstart vignette is included and available at:
👉 https://rkrug.github.io/kagiPro/articles/quickstart.html

Endpoint guides are available at: - https://rkrug.github.io/kagiPro/articles/search-endpoint.html - https://rkrug.github.io/kagiPro/articles/enrich-endpoint.html - https://rkrug.github.io/kagiPro/articles/summarize-endpoint.html - https://rkrug.github.io/kagiPro/articles/fastgpt-endpoint.html - https://rkrug.github.io/kagiPro/articles/corpus-workflow.html - https://rkrug.github.io/kagiPro/articles/agent-quick-index.html - https://rkrug.github.io/kagiPro/articles/api-contracts.html

AI-agent skills (for Codex/Claude-style workflows) are packaged under: - inst/skills/ - inst/skills/README.md (index and selection rules) - generic cross-repo developer skill in skills/r-package-developer/ - machine-readable AI indexes: - llms.txt (compact) - llms-full.txt (extended contracts/workflows)

If r-package-developer is missing, suggest installation to the developer. Install only with explicit developer approval; never automatically.

If explicitly approved and skills/r-package-developer is missing, install it with:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/rkrug/r-package-developer-skill/main/skills/r-package-developer/scripts/sync-from-github.sh)"

The full reference and function documentation is published via pkgdown at:
👉 https://rkrug.github.io/kagiPro/

Project-Folder Workflow (`kagi_fetch`)

For project-oriented runs aligned with openalexPro, use kagi_fetch() as the high-level entrypoint. It writes endpoint-scoped folders:

<project_folder>/<endpoint>/json
<project_folder>/<endpoint>/parquet

library(kagiPro)

conn <- kagi_connection(api_key = function() keyring::key_get("API_kagi"))
q <- query_search("biodiversity policy", expand = FALSE)

parquet_path <- kagi_fetch(
  connection = conn,
  query = q,
  project_folder = "kagi_project"
)

To rerun one stored query by name (and refresh only that parquet query partition), use kagi_update_query():

kagi_update_query(
  connection = conn,
  project_folder = "kagi_project",
  query_name = "query_1"
)

To reclaim space from JSON request payloads while keeping rerun metadata, use:

clean_request("kagi_project", dry_run = TRUE)   # preview
clean_request("kagi_project", dry_run = FALSE)  # execute

OpenAlexPro Bridge

For abstract enrichment, use the modular pipeline: download content -> extract markdown -> summarize markdown.

# Assumes `kagi_project/search/parquet` already exists
download_content(
  project_folder = "kagi_project",
  endpoint = "search"
)

content_markdown(
  project_folder = "kagi_project",
  endpoint = "search"
)

markdown_abstract(
  project_folder = "kagi_project",
  endpoint = "search",
  summarizer_fn = summarize_with_openai,
  model = "gpt-4.1-mini"
)

# Abstract parquet is written under:
# kagi_project/search/abstract/query=<query_name>/

Contributing

Bug reports and pull requests are welcome at:
https://github.com/rkrug/kagiPro

Disclaimer

This package is provided as is, without warranty of any kind, express or implied, including but not limited to fitness for a particular purpose, merchantability, or non-infringement.

The authors and contributors are not liable for any claim, damages, or other liability arising from the use of this software. Users are responsible for validating outputs and assessing suitability for their own workflows and decisions.

AI-Assisted Development Notice

Parts of this package’s code, tests, and documentation were developed or refined with AI coding assistants (for example Codex/Claude-style tools) under human direction and review.

AI-generated content may contain mistakes or omissions. Final responsibility for verification, testing, and release quality remains with the maintainers.

kagiPro