Documentation

Start with local document versioning.

RagTrack gives RAG teams a small CLI-first layer for ingesting documents, creating versions, chunking content, searching latest knowledge, and checking diffs.

Quickstart

Install

pip install ragtrack

Create docs

mkdir docs
cp house_rules.pdf docs/

Ingest

ragtrack ingest ./docs

Search

ragtrack search "wifi password"

CLI Commands

ragtrack ingest ./docs

Scans supported files, creates versions, chunks content, and stores everything in SQLite.

ragtrack status

Shows total documents, versions, chunks, and the latest version for each document.

ragtrack diff

Compares latest document versions against their previous versions using chunk hashes.

ragtrack search "query"

Searches latest chunks. Keyword mode works by default; semantic mode uses an embedding extra.

ragtrack serve

Starts the local web interface for uploads, browsing, search, diffs, and settings.

Search Modes

Keyword search is the default and works without embedding dependencies. Semantic search is optional and uses FastEmbed, a CPU-friendly ONNX Runtime embedding provider.

pip install "ragtrack[embed]"
SEARCH_MODE=semantic ragtrack ingest ./docs
ragtrack search "late checkout policy"

Docker

Use Docker when you want the same lightweight environment on a VPS.

docker compose run --rm ragtrack ingest /docs
docker compose run --rm ragtrack search "wifi password"
docker compose run --rm ragtrack diff
docker compose run --rm ragtrack status

Environment

SEARCH_MODE keyword

Use keyword or semantic search.

EMBEDDING_PROVIDER fastembed

Embedding provider used when semantic search is enabled.

RAGTRACK_DATA_DIR .ragtrack

Where SQLite and local index metadata are stored.

RAGTRACK_PROJECT_ROOT current directory

Root directory RagTrack uses for local project state.

Use With Larger Systems

Put RagTrack between your source documents and your AI application. Your assistant uploads or syncs files into RagTrack, RagTrack versions and chunks them, then the assistant searches the latest chunks before answering.

Apartment PDFs -> RagTrack ingest -> latest chunks -> AI assistant answer
Changed house rules -> RagTrack v2 -> diff + search latest version