How does recursive summarization handle huge documents?

It splits the text into chunks that each fit the context window, summarizes every chunk in parallel, then concatenates those summaries and summarizes them again. Repeating this map-reduce collapse means a document of any length eventually reduces to a single summary, no matter how far it exceeds the model's context limit.

Where does my API key go?

It stays in your browser and is sent only in the direct requests to OpenAI or Anthropic. It is never stored or sent to any Gera server. Every chunk and reduction call goes straight from your browser to the provider.

How many API calls will this make?

One call per chunk in the first pass, then one call per group in each reduction pass. A 100,000-character document at 6,000 characters per chunk makes about 17 first-pass calls plus a few reduction calls. Larger chunks reduce the call count but use more tokens per call.

Some compression is unavoidable — that is the point of summarization. Recursive summarization preserves the high-level structure well but drops fine detail in deeper passes. Use a larger chunk size and fewer passes when you need to retain more nuance.

Recursive Summarizer (BYO-key)

Summarize documents that do not fit in context

Even large context windows have limits, and feeding a 200-page report to a single call is slow and expensive. Recursive summarization solves this with a map-reduce approach: summarize each chunk, then summarize those summaries, repeating until one coherent summary remains. This tool does it in your browser with your own OpenAI or Anthropic API key.

How it works

You paste the document, set a chunk size, and provide your key. The tool splits the text on paragraph and sentence boundaries into chunks of roughly your chosen size, then summarizes each chunk with a real API call. It concatenates the chunk summaries and, if they are still long, summarizes them again — repeating this reduction until a single summary is produced. Progress is shown as each layer completes, and the final summary plus the intermediate chunk summaries are displayed. Every call goes directly from your browser to the provider.

Tips for better summaries

Tune chunk size to the trade-off. Larger chunks preserve more context per summary but cost more tokens per call; smaller chunks are cheaper per call but lose cross-chunk context.
Use a cheap model for the first pass. GPT-4o mini or Claude Haiku handle chunk summarization well; reserve a stronger model for the final reduction if you need a polished result.
Clean the input first. Stripping boilerplate, headers, and footers before summarizing improves quality and cuts cost.
Expect some detail loss in deep passes — that is inherent to compression.