Email Search CLI

Process email archives (PST files) into a searchable ChromaDB vector database with automatic semantic embeddings. Ingest, classify, search, analyse, and export to markdown.

Prerequisites

• Python 3.11+
• Package installed in a virtual environment

First-Time Setup

# Set up Python environment (one-time)
~/.claude/skills/email-search/setup.sh

Usage

All commands use the email-search CLI, which should be run from the skill's virtual environment:

~/.claude/skills/email-search/.venv/bin/email-search <command>

Ingest a PST file

email-search ingest /path/to/archive.pst

Emails are classified into three tiers:

• Tier 1 (Excluded): Calendar invites, delivery notifications, password resets - skipped
• Tier 2 (Metadata Only): Short emails, automated senders - stored but not vectorised
• Tier 3 (Vectorise): Real conversations - stored and auto-embedded

Search emails

email-search search "budget meeting notes"
email-search search "project update" --from 2023-01-01 --to 2023-12-31
email-search search "invoice" --sender "accounting"
email-search search "contract terms" --emails-only
email-search search "spreadsheet data" --attachments-only --limit 5

Analytics

email-search analytics summary
email-search analytics timeline
email-search analytics timeline --monthly --year 2020
email-search analytics contacts --limit 50

Export to markdown

email-search export contacts -o areas/relationships/email-contacts.md
email-search export review -p week -d 2023-01-15
email-search export review -p month -o reviews/monthly/2023-01-email.md

Status

email-search status

Configuration

# Default data directory: ./email-search-data
export EMAIL_SEARCH_DATA_DIR="/path/to/data"

Key Details

• No API keys or external services needed - ChromaDB runs locally with built-in embeddings (all-MiniLM-L6-v2)
• Text extraction from PDF, Word, Excel, and plain text attachments
• No data leaves the machine