condense-json 1.1

Simon Willison's Blog 08/03/26, 04:56 AM Tools

Summary

condense-json 1.1 adds structural replacements and merge operations for objects, plus property-based round-trip tests using Hypothesis.

No content available

Original Article

View Cached Full Text

Cached at: 08/05/26, 01:39 AM

# Release: condense-json 1.1 Source: [https://simonwillison.net/2026/Aug/3/condense-json/](https://simonwillison.net/2026/Aug/3/condense-json/) After shipping[condense\-json 1\.0](https://simonwillison.net/2026/Aug/2/condense-json/)I started integrating it into LLM, and found there were some desirable new features already: > - Replacements object can now include values other than strings\. These will be identified and used as structural replacements by`condense\_json\(\)`and`uncondense\_json\(\)`\.[\#8](https://github.com/simonw/condense-json/pull/8) - Objects can be used as the basis for merge operations\.`condense\_json\(\)`will identify if there are objects that are a close match and will store instructions for keys to update or delete\.`uncondense\_json\(\)`can then apply these merges\. I also added[some round\-trip tests](https://github.com/simonw/condense-json/blob/1.1/tests/test_properties.py)using the[Hypothesis](https://hypothesis.readthedocs.io/)property\-based Python testing library\.

Similar Articles

condense-json 1.0

Simon Willison's Blog

Simon Willison announces the 1.0 release of condense-json, a Python library that condenses JSON by replacing repeated substrings with compact references, saving space in SQLite logs for LLM.

TSON – A JSON superset with immutable, hash-pinned schemas

Hacker News Top

TSON is a schema system and JSON superset with immutable, hash-pinned schemas whose definitions are themselves data. It offers a compact schema grammar and a working Java implementation.

Context Compression Is Not One Thing: Readable Symbolic Re-expression vs. Coherent Summary at Matched Budget

arXiv cs.CL

This paper proposes Telegraph English, a readable symbolic format for context compression that outperforms matched-budget baselines on multi-hop QA datasets, preserving entity content more densely.

datasette 1.0a35

Simon Willison's Blog

Datasette 1.0a35 alpha release introduces a new create table interface and alter table JSON API, along with stable template context documentation.

@hasantoxr: Now turn messy documents into structured knowledge with one command. It's called Hyper-Extract. Most RAG tools just chu…

X AI KOLs Timeline

Hyper-Extract is a CLI tool that transforms messy, unstructured documents into structured knowledge such as knowledge graphs, hypergraphs, temporal/spatial graphs, and Obsidian vaults, supporting local LLM inference and MCP integration.