jsonRepair

A lightweight R package to automatically fix malformed JSON strings—especially those produced by Large Language Models (LLMs).

"Some LLMs are a bit iffy when it comes to returning well‑formed JSON data—sometimes they skip a parenthesis, sometimes they add an extra word. Luckily, the mistakes LLMs make are simple enough to be fixed without destroying the content… I searched for a lightweight R package that was able to reliably fix this problem but couldn't find any. So I copy one."

✨ Key Features

• Self‑healing JSON – Detects and repairs the most common structural issues: missing/extra commas, mismatched brackets, single quotes, comments, unescaped control characters, and more.
• jsonlite friendly – Optionally returns an R object via jsonlite::fromJSON() or the repaired JSON text.
• Zero heavy dependencies – Only needs base R and jsonlite.

🧑‍💻 Why This Package?

1. LLMs love to bend the JSON spec. If you call an LLM from R and ask it to emit JSON, chances are you'll face dangling commas, stray quotes, or plain text mixed in the payload.
2. Existing solutions are Python‑only. The excellent json_repair library inspired this work, but I needed the same logic in native R for seamless data pipelines.
3. One‑to‑one logic port. This package is a line‑by‑line re‑implementation of the original algorithm, translated from Python to R.

🚀 Installation

# install.packages("devtools") if you haven't yet
# install directly from GitHub
remotes::install_github("cgxjdzz/jsonRepair")

⚡ Quick Start

library(jsonRepair)

broken <- '{"number": 1,"reason": "According...""ans": "YES"}'

# Return as an R list (default)
repaired_obj <- repair_json(broken)
print(repaired_obj)
#> $number
#> [1] 1
#> $reason
#> [1] "According..."
#> $ans
#> [1] "YES"

# Or get the repaired JSON string
repaired_txt <- repair_json(broken, return_objects = FALSE)
cat(repaired_txt)
# {"number":1,"reason":"According...","ans":"YES"}