Intro - Parsica Weave

​

At a glance

• Automatic capture - every turn chunked, tagged (files, commands, tools, session, time), scrubbed, and stored; nothing to remember to do
• Scrub before store - credentials and PII redacted at a single tested write gateway no code path can skip
• Retrieval over dumping - query --grep returns the right three chunks instead of reloading a conversation
• Working set, journal, lessons - a living per-project scratchpad plus workspace-level history and hard-won lessons, re-injected at every session start
• Compaction survival - complete turns flushed before compaction, orientation re-injected after; the full record stays queryable forever
• Semantic enrichment - topics, summaries, search queries, atomic facts, and decisions with conviction layered over the raw mirror
• Refusal and human-only lists - your own never-to-appear terms, managed by CLI or slash command, fenced everywhere content reaches the store
• Channels and isolation - named session lanes, plus PARSICA_ISOLATED=1 for full recall with zero footprint
• Local-first, stdlib-only - no pip dependencies, no network calls, no telemetry; plain-text files under ~/.claude/parsica-weave/ that are yours
• Retention rotates, never deletes - aged shards move to archive/, still queryable; no code path unlinks user data

​

Quick start

# 1. install + enable the parsica-weave plugin, then verify:
parsica-weave doctor

# 2. work a few turns in any project, then ask your history:
parsica-weave query --grep "something you said"

# 3. name your agent (optional, recommended):
parsica-weave name --name Aria