txi

A command-line tool for building a searchable index of conversation transcripts (Otter today, Slack later).

txi walks a directory of transcript files, parses them into a structured SQLite index, and exposes that index through full-text search, metadata filters, and (eventually) an MCP query surface.

Installation

Requires Python 3.12+.

Install from source with uv:

git clone https://github.com/aaronwolen/transcript-indexer.git
cd transcript-indexer
uv sync

How it works

Transcripts live as plain files on disk. txi sync reads each file, hashes its contents, and upserts conversations and turns into a local SQLite database with an FTS5 index. Repeated syncs are incremental: unchanged files are skipped, renames are detected via content hash, and orphaned rows are pruned.

Once synced, txi search and txi conversations query the index entirely locally.

Learn more

• Usage guide
• API reference