Adds z_track_start/z_track_stop + z_record to zhelpers.sh and wires
z_track_start into every command script. Each run now captures its own output
volume (FIFO+tee, ANSI stripped), prints the '--- N lines / N chars / ~N tokens
est. (Claude Code) ---' footer, and appends one JSONL row per top-level run in
the same shape as the PowerShell tokens.jsonl. A nested-run guard keeps
zrestart from double-counting its zkill/zstart children.
Data dir precedence: $ZTOKENS_DATA, config ztokens.dataDir, sibling
../../ztokens/data, else ~/.ztokens/data. Docs + example config updated.
Verified on WSL (isolated data dir): single run records correctly; nested
zrestart produces one combined record, not three; est = round(chars/3.5).
zstart's venv detection fell back to .venv/Scripts/python.exe (a Windows
binary) whenever it existed. On a Windows-built project accessed from WSL that
file sits on the mount and looks executable, so it got picked and failed with
'exec format error' instead of falling through to python3. Guard that branch to
Windows-family shells (msys/cygwin), where a .exe can actually run.
Verified on WSL: zstart --detached now backgrounds a stdlib app via python3,
serves HTTP 200, logs to /tmp/zstart-<key>.log, and zkill terminates it and
frees the port.
Full port: zhelpers.sh library (jq config, ssh, http/tcp, archive builder,
build-version, motd, port-kill), all commands (zstart/zkill/zrestart/zstop,
zdeploy with 5 kind handlers, zec2/zec2online/zrepair, zbackup/zbackup_ec2/
zsync/zbackup_and_sync, zstart_docker, zsetup_mail, setup_backup_schedule via
cron), Unix-path zconfig.example.json, and a bash/README.md.
Verified: bash -n clean on all scripts; archive exclusions, config semantics,
and zec2 tested against the real config and live server from Git Bash. Needs a
Linux/macOS/WSL shakedown for lsof/rsync/nohup paths before merging.
Per-run captures are measured; the ~26,500/day total multiplies them by assumed
typical run counts (zdeploy/zrestart at 10-15/day dominate). Label that boundary
explicitly in README, ELEVATOR_PITCH, and TOKEN_SAVINGS so the daily figure isn't
read as a direct measurement.
- One headline number everywhere: ~26,500 measured tokens/day (README said 3,000-7,000; ELEVATOR_PITCH said 157,000-540,000)
- Measurement note: measured output volume, estimated tokenization; /3.5 is conservative for code-heavy output
- Raw-orchestration column explicitly labeled an upper bound, not a prediction
- Measured dollar column priced at input rates (blended kept for est-raw only); note on re-sent context tokens
- Untrack md/Z-ScriptTokenData.md (superseded internal notes; md/ gitignored)
Config-driven PowerShell scripts to run infrastructure tasks (deploy, restart, backup, diagnostics) yourself instead of having an AI agent orchestrate them, to save agent tokens. Environment specifics live in zconfig.json (gitignored).
