# ranxianglei/billion-context

> A context-compression proxy plugin between a coding agent and the model API: folds conversation history into layered summaries, saves tokens, and lets a single session run for months within a small context window.

- Magnitude: 2.3 out of 10 — Steady
- Stars: 742 total · +7 stars measured 2026-10-10, 11:30–12:58 UTC, ≈ 19 by evening
- Star trust: star growth looks organic
- Category: AI agents · Language: TypeScript · Created: 2026-08-06 · Last push: 2026-10-10
- GitHub: https://github.com/ranxianglei/billion-context · Page: https://gitnova.dev/en/r/ranxianglei/billion-context

## Useful for

- Run the proxy between a coding agent and the model API so a long session does not hit the context window
- Fold conversation history into layered summaries via the compress tool and restore exact details via decompress
- Search compressed summaries and visible messages for a needed fragment via search_context

## Why it’s here

- Star-counter measurements on 2026-10-10 (UTC), 11:30–12:58: 735 → 742 stars (+7). This is the change over that interval.
- Estimated end-of-day forecast: about +19 stars, using observed gains and the previous day.
- GitHub Trending this week: #11, +246 stars.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 73
- Issues and pull requests: 2,672
- Watchers: 2
- Average over the last week: 35 per day
- Usual pace: 19 per day
- Stars in the last hour (measured): 5
- Latest release: v0.1.191 (2026-10-10)

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-09-11 … 2026-10-10: 5, 3, 6, 11, 8, 12, 12, 6, 9, 11, 21, 16, 8, 15, 15, 18, 29, 22, 26, 32, 28, 27, 27, 36, 31, 29, 50, 34, 44, 7

## Spotted in now

- GitHub Trending this week: #11, +246 stars

## More in this category

1. **morluto/rea** — 9.9 · Breakout · AI agents · TypeScript · +15,629 stars measured 2026-10-10, 00:31–12:54 UTC, ≈ 27,546 by evening
   A reverse-engineering tool that lets AI agents inspect apps and native binaries: it connects agents to Hopper or Ghidra via MCP and CLI, helping explain how a feature works and rebuild it in your own project.
   Full card: https://gitnova.dev/en/r/morluto/rea.md
2. **zhongerxin/iPhone-use** — 7.6 · Breakout · AI agents · Python · +670 stars measured 2026-10-10, 00:31–12:54 UTC, ≈ 1,235 by evening
   Lets Codex control a real iPhone over USB via WebDriverAgent: it installs and signs WDA, exposes MCP tools for app navigation, taps, typing and list collection, and shows a live screen widget.
   Full card: https://gitnova.dev/en/r/zhongerxin/iPhone-use.md
3. **Ebony-Vinyl/dsh-our-free-model** — 7.4 · Breakout · AI agents · JavaScript · +901 stars measured 2026-10-10, 00:31–12:54 UTC, ≈ 1,599 by evening
   A dsh plugin that adds free access to Muse Spark 1.3, MiMo V2.6 and other models without sign-up or API key, plus a local OpenAI-compatible proxy and usage dashboard.
   Full card: https://gitnova.dev/en/r/Ebony-Vinyl/dsh-our-free-model.md
4. **franzenzenhofer/big-arrow-on-the-screen** — 7.0 · Early signal · AI agents · Swift · +93 stars measured 2026-10-10, 00:31–12:54 UTC, ≈ 240 by evening
   A macOS CLI and agent skill that draws big arrows, boxes and labels on top of the screen to point a human at a specific element, without stealing focus and disappearing on its own.
   Full card: https://gitnova.dev/en/r/franzenzenhofer/big-arrow-on-the-screen.md
5. **nano-muse/nanoMuse** — 6.9 · Early signal · AI agents · Swift · +176 stars measured 2026-10-10, 00:31–12:54 UTC, ≈ 304 by evening
   Open-source personal AI agent for all your devices: performs tasks via shell, browser, MCP and screen control, remembers you and asks permission before irreversible actions. Runs on Android, iOS, desktop and browser, can be self-hosted.
   Full card: https://gitnova.dev/en/r/nano-muse/nanoMuse.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-10-10 13:04 UTC, updated every 30 minutes.