Skip to content

Releases: nextbridgehq/repocode

Release list

v0.1.0 — Initial Public Release

v0.1.0 — Initial Public Release Latest
Latest

Choose a tag to compare

View all tags
@nextbridgehq nextbridgehq released this 06 Aug 14:18
v0.1.0
fd5abb3

RepoCode packs a repository's source code into a token-budgeted context file for AI coding assistants — ChatGPT, Claude, Gemini, and others. It ranks files by structural importance, import relationships, and recency, then fits the highest-ranked files into your token budget through compression rather than silently dropping them.

This is the first public release, published as three packages:

Package Install
repocode npm install -g repocode
@repocode/core npm install @repocode/core
@repocode/mcp npx -p @repocode/mcp repocode-mcp

Highlights

Pipeline

  • File discovery + ranking by structural importance, import centrality, recency, and keyword density
  • Token-budget fitting with a compression cascade: tier1 (strip comments) → tier2 (strip function bodies) → tier3-llm (LLM semantic summary) → partial (trim to fit)
  • Import tracing across TypeScript, JavaScript, Python, Go, Rust, Ruby, Java, C, C++, PHP, and C#
  • Output as XML, Markdown, JSON, or plain text

CLI

  • Git-aware modes: --diff, --staged, --branch, --pr, --log, --range
  • Query-aware re-ranking (--query) and a built-in instruction library (--instruction review|bugfix|security-audit|...)
  • Interactive TUI for hand-picking files, and --watch mode for continuously updated context
  • --explain to inspect ranking and inclusion decisions before packing

MCP Server

  • Exposes pack_repository, search_files, get_file_context, get_git_context, and get_architecture as MCP tools for editor/assistant integration

Security

  • Secret scanning with redaction before any LLM call, configurable per security mode (warn / exclude / redact / fail)
  • Repository config execution blocked by default in MCP mode
  • Runs entirely locally unless tier3-llm compression or semantic query mode is explicitly enabled

Reliability

  • Consistent stderr warning/error reporting throughout the pipeline
  • Exponential backoff with jitter on LLM compression retries
  • Lock-protected, atomic cache writes safe against concurrent repocode processes
  • Backed by 400+ automated tests across all three packages

Requirements

Node.js ≥ 22

Links

README · Changelog · Issues

Assets 2
Loading