SeedPass/memory-update.md

Protocol Research Findings: The codebase uses bip_utils and nostr_sdk correctly. However, BIP85 implementation uses app_no=2 for symmetric key derivation, which is non-standard (usually 128 or 0). A regression test src/tests/test_bip85_compliance.py was added to document and monitor this behavior.

Memory Update - 2026-02-26

What was done

• Ran TORCH update and doctor checks for this repository.
• Preferred dedicated torch/ install path first, then fell back to repo-root updater flow.

Key findings

• npx --prefix torch --no-install torch-lock update --force failed with:
  • src and dest cannot be the same /home/user/Documents/GitHub/SeedPass/torch/src
• Repo-root flow succeeded:
  • npx --no-install torch-lock update --force
  • npx --no-install torch-lock doctor reported 7 passed, 0 warned, 0 failed.

Operational note

• In this repo, when torch-lock update is run with --prefix torch, path resolution can collide on torch/src.
• If that occurs, running the updater from repo root is a reliable fallback.

Resolution Applied - 2026-02-26

• Implemented same-path guards in torch/src/ops.mjs:
  • copyDir(src, dest) now skips when path.resolve(src) === path.resolve(dest).
  • copyFile(src, dest, ...) now skips no-op self-copy operations for identical resolved paths.
• This aligns with upstream TORCH fix notes for self-hosted installs where package root equals install target.

Validation After Fix

• npx --prefix torch --no-install torch-lock update --force now succeeds (no src and dest cannot be the same error).
• Update created backup: torch/_backups/backup_2026-02-26T13-45-30-661Z.
• npx --prefix torch --no-install torch-lock doctor reports: 7 passed, 0 warned, 0 failed.

Operator Context Note (Adam)

• bitvid is a different project Adam is working on.
• If a prompt context references bitvid instead of SeedPass, treat it as a likely working-directory mismatch.
• In that case, ask Adam to confirm whether the active directory/repo is correct before making changes.