moeix 0.2.2

High-performance trigram code search for humans and AI agents.
Documentation
moeix-0.2.2 has been yanked.

ix 🔍

High-signal code retrieval for humans and AI agents.
Sub-millisecond search across multi-gigabyte codebases with constant memory overhead.

License: MIT Rust PRs Welcome


⚡️ The Hero Visual (Proof of Life)

# Instant search across a massive repo
$ time ix "ConnectionTimeout" src/
[ix] managed by ixd (Status: idle)
src/network/client.rs:42:1280:    pub timeout: ConnectionTimeout,
src/config/defaults.rs:15:450:    pub const DEFAULT_TIMEOUT: ConnectionTimeout = 30s;

real    0m0.004s
user    0m0.002s
sys     0m0.002s

# Streaming compressed logs without OOM
$ ix -z "ERROR" logs/2026-03-29.gz
logs/2026-03-29.gz:1042:0: [ERROR] Database connection failed

🎯 The Hook

ix is a high-performance, Unix-native search engine that eliminates the linear scan bottleneck of traditional tools. By leveraging a sparse trigram index and a constant-memory streaming architecture, ix delivers sub-millisecond retrieval across multi-gigabyte codebases with guaranteed resource safety.

Engineered for the 2026 agentic workflow, it introduces the Beacon Protocol to eliminate redundant re-indexing and provide authoritative state for AI agents.


🚀 Quick Start

Installation

cargo install --path . --features full

Build your first index

ix --build

This creates a compact .ix/shard.ix file (typically <10% of source size).


💎 Features

  • Universal Search: Seamlessly query .rs, .py, .gz, .zst, .zip, and .tar.gz files.
  • Beacon Protocol: Coordination plane between CLI and Daemon. Eliminates "stale index" confusion for AI agents.
  • Streaming Architecture: Search arbitrarily large files with constant memory overhead.
  • Data Integrity: Per-posting-list CRC32C checksums detect silent data corruption.
  • Logical Correctness: Full support for CRLF (\r\n) and LF (\n) line endings with accurate byte offsets.
  • Agent-Native: First-class support for --json output and context capping to respect LLM window limits.

🤖 Agentic Retrieval (LLM Usage)

ix follows the UTCP Schema for optimal AI agent integration:

Command Capability Agent Output
ix -c "p" Existence Check Single integer (match count)
ix -l "p" Location List of unique file paths
ix -C 3 "p" Contextual Retrieval ±3 lines around every match
ix --json "p" Structured Extraction JSON Lines for machine parsing
ix -r -U \"foo.*\\nbar\" Multiline Cross-line Regex support

🛠 Debugging & Development

For contributors and power users:

  • Telemetry: Use --stats to see retrieval performance and index efficiency.
  • Integrity: ix automatically verifies checksums during query execution.
  • Diagnostics: cargo test runs the full suite, including Negative Oracle corruption tests and Resource Boundary memory tests.

🏛 Architecture

ix employs a tiered verification pipeline:

  1. Query Planner: Analyzes Regex/Literal and expands trigram variants.
  2. Sparse Lookup: Intersects posting lists from the trigram table.
  3. Beacon Gate: Validates authority and coordinates with the background daemon (ixd).
  4. Streaming Verification: Byte-level verification of candidate matches using constant-memory buffers.

📜 License & Provenance


"Stop scanning. Start finding."