Back to the chart

ranxianglei/billion-context

Context-compression plugin for long coding sessions

Chart
Not on a chart right now
Stars
7.9k

Open on GitHub

What it is

A proxy and plugin that sits between coding agents and model APIs to rewrite Anthropic and OpenAI streams with incremental context compression. Instead of hard truncation, the model calls built-in tools to compress, decompress, and search conversation history.

Who it is for

Developers using AI coding agents who want to run long single sessions without hitting context window limits or high token costs.

Why it is trending

It gained 6,240 stars today to reach 7,741 total stars alongside the v0.1.216 release on November 17, 2026. It stands out for incremental, reversible compression that preserves prefix-cache hit rates.

Main features

  • Model-driven context management via compress, decompress, search_context, and acp_status tools
  • Incremental, reversible compression that maintains 96-99% prefix-cache hit rates
  • Supports native plugins, a zero-config launcher (bili <client>), and /bili/ URL prefixing
  • Works with coding clients including pi, opencode, claude, codex, aider, and zed

Summarized on November 17, 2026 from the README and repository data. It can be wrong; check the source before relying on it. Report a wrong summary