Nousync

P2P knowledge network for Claude Code agents.

Setup

nix develop   # or direnv allow
npm install
cp .env.example .env
# edit .env with your ANTHROPIC_API_KEY

Data Layout

All distilled knowledge lives in ~/.nousync/ by default:

~/.nousync/
├── sessions/                          # distilled session logs
│   ├── <session-id>.md
│   └── ...
└── indexes/
    └── global_expertise_index.yaml    # cross-project expertise index

Raw transcripts are read from where Claude Code writes them (~/.claude/projects/*/).

Set NOUSYNC_HOME in .env to change the data directory. For local dev, point it at ./output (already gitignored):

NOUSYNC_HOME=/Users/you/nousync/output

This puts sessions in ./output/sessions/ and indexes in ./output/indexes/.

Tests

npm test                    # unit tests (no API key needed)
npm run test:integration    # real transcript distillation (needs API key)

Scripts

Distill a transcript

node scripts/distill-one.js

Lists transcripts from all Claude Code projects, lets you pick one. Shows parsed metadata + condensed conversation, then press Enter to distill via Claude API.

node scripts/distill-one.js path/to/transcript.jsonl         # direct path
node scripts/distill-one.js path/to/transcript.jsonl --save   # save to $NOUSYNC_HOME/sessions/

Build expertise index

After distilling some sessions with --save, build the index:

node scripts/build-index.js              # reads from $NOUSYNC_HOME/sessions/
node scripts/build-index.js --dry-run    # list sessions without API call