Skip to content

Loki Mode

v7.32.0 Breaking

This release includes 3 breaking changes for platform teams planning a safe upgrade.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai-agents aider anthropic autonomous ci-cd claude
+13 more
cline code-review devops gemini github-action github-issues loki-mode multi-agent openai-codex openapi pull-request-review sdlc spec-driven-development

Summary

AI summary

Updates json/csv/timeline, heal/migrate, and explain/onboard/code/context across a mixed release.

Full changelog

Added

  • CLI consolidation Phase B: loki report kpis is the canonical KPI command
    (old loki kpis kept as a working alias); loki memory compound,
    loki analyze (explain/onboard/code/context), and loki modernize
    (heal/migrate) nouns group their legacy commands under the alias contract:
    byte-identical stdout, exit parity, exactly one stderr deprecation pointer,
    suppressed under --json/-q/--quiet and positional machine formats
    (json/csv/timeline), zero side effects in clean directories.
  • Homebrew now installs from a counted GitHub Release asset: the
    update-homebrew release job builds a git archive tarball, uploads it to
    the GitHub Release, and points the formula at it (sha256 of the same
    bytes), so brew installs appear in release download counts.
  • tests/test_dashboard_port_arg.py: coverage for the v7.31 --port/--host
    argparse on direct python -m dashboard.server launches.

Fixed

  • Cost-honesty (task 568): on the stock session path the estimator and the
    dashboard now quote the model the runner actually dispatches (a sonnet
    session pin routes through the development tier to Opus; the quote said
    Sonnet and underquoted ~1.7x). loki plan output now names the model AND
    the lever that selected it. Demo cost quote corrected to the Opus truth.
  • Tier-name session pins (planning/development/fast), documented in
    skills/model-selection.md, now resolve byte-faithfully across all three
    readers (run.sh dispatch, loki plan estimator, dashboard). run.sh folds
    miscased/padded pins (trim + lowercase). Session-pin parity matrix grown
    to 224 cells; dashboard endpoint tests 37/37.
  • loki report export json kpis was hijacked by a flag-anywhere kpis scan;
    kpis routing now keys on the first non-flag token in all three layers
    (bash dispatch, bin/loki shim, Bun CLI) with byte-parity against main.
  • loki mcp probe now resolves the server exactly like the launch exec
    (file-exec of mcp/server.py with PYTHONPATH from the user cwd), so probe
    and launch can no longer disagree under a decoy SDK; the runpy
    RuntimeWarning on launch is gone and MCP serverInfo reports the real
    loki-mode version.
  • A backtick command substitution inside the loki plan estimator comment
    block executed a stray claude --model on every plan invocation,
    polluting stderr (plan suite 19/25). Fixed; loki plan --json now emits
    zero stderr bytes (25/25).
  • Homebrew formula license corrected from MIT to BUSL-1.1 (honesty fix) and
    description updated.

Changed

  • loki compound help title reads the canonical loki memory compound
    (the alias keeps working with a pointer).
  • scripts/local-ci.sh MCP handshake check launches the server in the shipped
    launcher form (file-exec + PYTHONPATH from a non-repo cwd) instead of
    -m mcp.server from the repo root.

Breaking Changes

  • Legacy command `loki kpis` retained as alias for new canonical `loki report kpis`
  • Legacy commands grouped under new nouns: `loki analyze` (explain/onboard/code/context) and `loki modernize` (heal/migrate)
  • Homebrew formula license changed from MIT to BUSL-1.1

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track Loki Mode

Get notified when new releases ship.

Sign up free

About Loki Mode

Multi-agent autonomous SDLC framework. Spec to deployed app. PRD, GitHub issue, OpenAPI/JSON/YAML, or one-line brief. 5 AI providers, 11 quality gates.

All releases →

Related context

Beta — feedback welcome: [email protected]