Skip to content

Loki Mode

v7.40.0 Breaking

This release includes 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

Affected surfaces

auth rbac

Summary

AI summary

Fixes complexity detection on bash route ensuring accurate analysis gating.

Full changelog

Added

  • Autonomous, complexity-gated decision for the no-PRD codebase-analysis pass.
    When no PRD is provided, Loki reverse-engineers a PRD from the existing code.
    On the Claude provider, when the repo is detected as complex (file_count > 50,
    microservices, external deps, or a complex inferred spec), Loki now dispatches
    that read-only analysis pass as a Claude Code Dynamic Workflow (parallel
    fan-out, the ultracode trigger) for more thorough first-run analysis, and
    otherwise keeps the cheaper three-pass analysis. The decision is made once per
    run (not always-on) and is parity-locked across both routes
    (autonomy/run.sh and loki-ts/src/runner/build_prompt.ts), evaluated in the same
    order: non-Claude provider or degraded provider -> three-pass;
    LOKI_USE_CLAUDE_WORKFLOWS=0 -> three-pass (escape hatch);
    LOKI_USE_CLAUDE_WORKFLOWS=1 -> workflow (force on); variable unset ->
    workflow only when complexity == complex. Simple and standard repos keep the
    byte-identical default behavior. When the workflow path is taken autonomously,
    a one-time stderr disclosure names the higher cost class (no fabricated dollar
    figure; there is no price API) and the LOKI_USE_CLAUDE_WORKFLOWS=0 opt-out.

Fixed

  • detect_complexity() was defined but never called on the bash route, so
    DETECTED_COMPLEXITY stayed empty and every complexity consumer
    (effort-for-tier, telemetry, phase selection, and the new analysis gate)
    silently fell back to "standard". It now runs once per run before the first
    prompt build, so the complexity signal is live on bash, matching the Bun route.
  • parity-drift CI check: removed an over-normalization that blanked summary
    counts, which was hiding real count drift between the bash and Bun routes.
    The legitimate disk-space and route-specific Runtime normalizations are kept,
    so genuine drift is caught while the prior false positive stays closed.

Docs

  • Installation guide now leads with Bun as the recommended install
    (bun install -g loki-mode) with npm as a fully supported alternative,
    matching the README and clarifying that the core autonomous engine runs the
    same on both routes (you lose no capability by choosing Bun).

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]