um-pifetch

um-pifetch fetches web content once into a local, content-addressed cache and searches that cache afterwards. Repeat lookups hit the local copy instead of the network.

Setup

The script is a self-contained uv script (Python, httpx dependency declared inline). Run it through uv directly or place it on PATH and execute it; uv resolves the dependencies on first run.

Usage

um-pifetch get <topic> <url> [--stable | --ttl T]   # fetch-or-cache-hit, print path
um-pifetch rg <topic> <pattern> [rg-args...]        # rg -z over the cached topic
um-pifetch status <topic-or-url>                    # age, sha256, policy
um-pifetch list [topic]                             # manifest summary
um-pifetch prune [--older-than T] [--all] [--dry-run]

• TTL syntax: integers with suffix d, h or m (for example 3d, 12h, 30m); the default is 3d.
• prune never drops --stable topics unless --all is passed; that flag is the user's approval.

The design is last-wins content with an append-only event manifest and conditional GETs. A companion skill document, fetched-search, describes the recommended lookup workflow.