goosedump

goosedump is a coding agent context data browser. It supports Claude, Codex, Crush, Gemini, Goose, OpenCode, and Pi. It builds on top of the ideas of lllyasviel/VCC.

Features:

• Native-format output on stdout, in each coding agent's own session shape.
• Format conversion: re-render any context in another provider's native format.
• BM25 keyword ranking for natural queries.
• Glob matching for filtering messages.

Usage

# List contexts across all providers (optionally filtered by a glob).
goosedump list
goosedump list goose:*

# Show a full transcript in the source provider's native format.
goosedump show goose:abc123

# Convert it to another provider's native format by appending a target format.
# JSONL backends (claude, codex, pi) emit the exact native session file;
# SQLite backends (crush, goose, opencode) emit a row-oriented JSON projection
# of their tables that is straightforward to insert into the live store.
goosedump show claude:abc123 goose

# Filter or rank messages, optionally converting the matches too.
goosedump grep claude:abc123 "TODO" goose
goosedump search claude:abc123 "race condition" -p2

# Produce a compaction summary (always emitted in its own fixed schema).
goosedump compact goose:abc123

A converted context round-trips: the output re-reads through the target provider's reader to the same internal representation. Session metadata the internal representation does not carry (per-message timestamps, model names) is omitted or synthesized.

Build

cargo build
cargo test

Licensing

The JavaScript npm wrapper is licensed under Apache-2.0. The Rust source and native binaries are licensed under LGPL-2.1-or-later. Corresponding source for each published native binary is available from the GitHub repository tag that matches the package version.