Internal crate for Vestige — embedding provider abstraction. Deterministic fake provider always available; fastembed and ollama land in PR7 behind feature flags.
Local-first, repo-pinned memory layer for coding agents. CLI + MCP server over a SQLite store. No daemon.
Internal crate for Vestige — `.vestige/config.toml` round-trip and project identity resolution. Public API is unstable; use the `vestige` binary.
Internal crate for Vestige — pure domain types, memory engine, representation derivation, ranking. Public API is unstable; use the `vestige` binary.
Vestige orchestration layer — hybrid search and embed ingest over Store + EmbeddingProvider.
Internal crate for Vestige — MCP server adapter (rmcp 0.16, stdio transport). Public API is unstable; use the `vestige` binary.
Internal crate for Vestige — SQLite (bundled, FTS5) persistence layer. Public API is unstable; use the `vestige` binary.
A command line tool to invoke and work with Large Language models on AWS, using Amazon Bedrock
Senior SysAdmin, Network Admin, Data Analyst, and Software Engineer living in your terminal. A high-precision local AI agent harness for LM Studio, Ollama, and other local OpenAI-compatible runtimes that runs 100% on your own silicon. Reads repos, edits files, runs builds, inspects full network state and workstation telemetry, and runs real Python/JS for data analysis.