Mailprune

Context & Motivation

Context: Personal inboxes accumulate noise over time; Mailprune emerged to audit Gmail accounts locally, identify high-volume low-value senders, and recommend targeted fixes.

Motivation: Large inboxes and API rate limits make full audits slow and fragile. Users need actionable recommendations that preserve privacy, avoid accidental mass changes, and can be reviewed before applying.

The Local Implementation

• Current Logic: Mailprune is a Python CLI that uses the Gmail API to fetch messages and metadata, computes sender-level statistics (volume, open-rate proxies, thread activity), clusters senders by content/behavior, and produces targeted recommendations (unsubscribe, filter, mute). It runs as a local audit: uv run mailprune audit pulls messages, analyzes them in parallel (thread/process pool), caches intermediate results, and writes reports to disk.
• Bottleneck: Fetching messages at API rate limits and large inbox sizes creates IO-bound runs; memory and CPU usage spike during full-content clustering and embedding computation; OAuth token lifecycle and reauth flows complicate long-running jobs; absent persistent checkpoints, interrupted runs must restart from scratch.

Comparison to Industry Standards

• My Project: Mailprune — local-first, privacy-aware audits + explainable recommendations; focuses on sender clustering and engagement heuristics.
• Industry (e.g., SaneBox / CleanEmail / Gmail filters): SaaS products offer managed cleanup with proprietary heuristics and cloud processing; Gmail filters operate at the server level and are immediate but require manual rule authoring.
• Gap Analysis:
  • Feature parity: SaaS can provide real-time, always-on filtering and cross-device integration; Mailprune is audit-first with manual or semi-automated remediation.
  • Cost: Building distributed ingestion, reliable background workers, and persistence raises engineering and infra costs compared to single-node local tooling.
  • Benefit: Mailprune’s privacy-first approach and audit transparency are differentiators for privacy-conscious users.

Risks & Mitigations

• OAuth/credentials leaks: encrypt tokens at rest; use OS keychain when available; rotate refresh tokens and surface reauth UX.
• Privacy violation from cloud embeddings: default to local embeddings; if cloud is enabled, show explicit consent and purge raw message text after embedding.