transcribe-it 0.2.0

Lightweight CLI for ingesting, enriching, and storing meeting transcripts

transcribe-it

A lightweight CLI for ingesting meeting transcripts (Gmail or Slack), enriching them with an LLM, and storing the results as local files.

Source -> Extract -> LLM Enrich -> Local Files

Prerequisites

• Python 3.12+
• A Google Cloud project with Gmail API + Google Drive API enabled (for the Gmail source), or a Slack bot token (for the Slack source)
• An API key for one of the supported LLM providers (Anthropic, OpenAI, or Groq) — only needed if you want LLM enrichment

Install

uv tool install transcribe-it

Or with pipx:

pipx install transcribe-it

Setup

Run the interactive setup from the directory where you want transcripts to be stored:

transcribe-it init

This will:

• Ask which sources to enable (Gmail, Slack)
• Prompt for the credentials each source needs
• Optionally let you pick an LLM provider and store the API key (skip this if you only want raw transcripts)
• Write .transcripts/config.yaml in the current directory
• Write secrets to ~/.config/transcript/env

Gmail credentials

For Gmail you'll be asked for GOOGLE_OAUTH_CLIENT_ID and GOOGLE_OAUTH_CLIENT_SECRET. Two options:

1. Reuse someone else's OAuth client — ask a teammate for the values and have them add your Google account as a Test user on their OAuth consent screen.
2. Create your own — in Google Cloud Console, create an OAuth 2.0 Client ID of type Desktop app, then copy the client ID and secret from the resulting credentials.

After init, authenticate:

transcribe-it auth gmail

Slack credentials

For Slack you'll be asked for the channel ID and a bot token (xoxb-...). The bot needs to be a member of the channels you want to ingest from.

Usage

By default, ingestion only extracts the raw transcript — no LLM call, no API key required. Pass --enrich to also generate a summary, topics, and participants via LLM.

# Last N days, raw extraction only (default)
transcribe-it ingest gmail --days 7

# With LLM enrichment
transcribe-it ingest gmail --days 7 --enrich

# Enrichment + cleaned transcript variant (--clean implies --enrich)
transcribe-it ingest gmail --days 7 --clean

# Specific date range
transcribe-it ingest gmail --from 2026-04-01 --to 2026-04-05

# Preview matching emails without fetching or writing
transcribe-it ingest gmail --days 1 --dry-run

# Ingest a single transcript file directly
transcribe-it ingest file path/to/transcript.txt

Output

Raw mode (default) writes a single .txt file per transcript:

.transcripts/
  2026-04-09-ai-labs-daily.txt

With --enrich, each transcript becomes a folder:

.transcripts/
  2026-04-09-ai-labs-daily/
    raw.txt          # Original transcript (immutable)
    metadata.json    # Source, date, participants, topics, summary

With --clean, an additional clean.md is written (structured: title, summary, topics, cleaned transcript).

Prompts

LLM prompts are bundled with the package under transcribe_it/prompts/. To customise, fork the repo and edit prompts/enrich.md.

Commands

Command	Description
transcribe-it init	Interactive setup for sources, credentials, and LLM
transcribe-it auth gmail	Authenticate with Gmail (OAuth)
transcribe-it ingest gmail	Ingest transcripts from Gmail
transcribe-it ingest file PATH	Ingest a single transcript file

Ingest options (Gmail)

Flag	Description
--days N	How many days back to search
--from YYYY-MM-DD	Start date
--to YYYY-MM-DD	End date
--profile NAME	Gmail auth profile
--dry-run	List matching emails without processing
--enrich	Run LLM enrichment (summary, topics, participants)
--clean	Also generate a cleaned version of the transcript (implies --enrich)

Configuration files

Path	Purpose
.transcripts/config.yaml	Per-project: sources, lookback, output destinations
~/.config/transcript/env	Global: API keys and OAuth credentials