Skip to content

Loki Mode

v7.45.0 Security

This release includes 1 security fix for security teams reviewing exposed deployments.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →
This release patches 1 known CVE

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

Affected surfaces

auth

Summary

AI summary

Docker images now build with a default AI provider CLI installed, fixing non‑runnable builds and patching an empty .env security issue.

Full changelog

Docker is now a first-class way to run Loki. The published image could not run a
single build (it shipped the runtime but no AI provider CLI), so this release
makes the image buildable, adds a zero-friction loki docker host wrapper, and a
docker compose .env workflow, plus a security fix for a tracked empty .env.

Fixed

  • Docker image could not run builds. The published asklokesh/loki-mode image
    shipped the Loki runtime (bun/node/python) but NO AI provider CLI, so
    docker run ... start died at the first provider call with "No AI provider CLI
    found; cannot prompt to install in a non-interactive shell." The Dockerfile now
    installs @anthropic-ai/claude-code (the Tier-1 default provider). Verified:
    claude --version (2.1.177) runs in the image and a claude -p call succeeds
    with valid auth.
  • Security: a tracked empty .env defeated gitignore. An empty .env had been
    committed since v6.36.2, so .gitignore could not protect a user's real .env.
    Untracked it (git rm --cached .env) and added .env, .env.*, and
    .loki-oauth-credentials.json to .gitignore (with !.env.example kept
    tracked).
  • Docker resume refused to restart in the same folder. Inside a --rm
    container the orchestrator is always PID 1, so a prior run recorded
    loki.pid=1. On the next loki docker start in that folder the stale-PID
    lock check ran kill -0 1, which always succeeds (PID 1 is the live init
    process), so the lock falsely read "another loki instance is running" and
    refused to resume. Fixed: a recorded pid of 1 from a prior container is now
    treated as stale unless it is the live container's own pid ($$ == 1), so
    re-running loki docker start in the same folder resumes correctly with full
    memory and session continuity.

Added

  • loki docker <any-loki-command> zero-friction host wrapper. Runs loki inside
    the published image with the same experience as the local CLI. It bind-mounts
    the current folder to /workspace so .loki/ state (memory, session, queue,
    checkpoints) persists on the host and resume + continuity work exactly like
    local. Auth is auto-detected: ANTHROPIC_API_KEY if set, else the host's
    existing Claude Code login (macOS keychain or ~/.claude/.credentials.json) is
    extracted to a private 0600 temp file and mounted read-write (so the
    in-container claude can refresh the short-lived token and it persists), else an
    honest error with setup guidance. Forwards ~/.gitconfig, ~/.config/gh, and
    GITHUB_TOKEN/GH_TOKEN. Flags: --image, --dry-run. New module
    autonomy/docker-run.sh with 24 hermetic unit tests (tests/test-docker-run.sh).
  • Multi-repo + unified-dashboard parity for loki docker. You can now run
    loki docker start in several different repos at once, exactly like the host
    CLI supports many repos. Each repo gets its own container (deterministic name
    loki-<sha256-prefix-of-path>), its own bind-mounted /workspace, and its own
    .loki/ state on the host, so two repos run as two concurrent containers
    without colliding. Each loki docker project also registers on the HOST with
    its real folder path and no pid into the same machine-global registry
    (~/.loki/dashboard/projects.json) the host dashboard reads, so a single host
    loki dashboard lists ALL your projects whether they run via local
    loki start or via loki docker start (the dashboard derives "running" from
    each project's bind-mounted .loki/session.json). Docker is now a first-class
    multi-repo surface with one unified view, matching the local CLI experience.
    Builds run with the dashboard off by default (so concurrent runs never fight
    over port 57374); use loki start --api or the host loki dashboard for the
    UI.
  • docker compose .env workflow. docker-compose.yml now has env_file: .env, an
    ANTHROPIC_API_KEY passthrough, and a commented host-OAuth-mount volume. New
    .env.example template. Users copy .env.example to .env, set their key, run
    docker compose run loki start prd.md, and edit/re-run without retyping flags.

Changed

  • Docker docs currency. Removed deprecated Gemini references from DOCKER_README.md
    and the release.yml Docker Hub short-description. Fixed stale Docker image tags
    (were 7.40.0) in README.md. Documented both auth methods and the loki docker
    wrapper across DOCKER_README.md, README.md, docs/INSTALLATION.md, and SKILL.md.

Security Fixes

  • .gitignore updated to ignore empty `.env`, `.env.*`, and `.loki-oauth-credentials.json` files (tracked issue from v6.36.2)

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]