ToggleLogic Free 1.0.2
ToggleLogic Free 1.0.2
ToggleLogic Free-Tier is source-available model routing for OpenClaw: sticky owner/session model overrides, static configured routes, a deliberately-simple cheapest-default, and a structured audit stream. Patent pending.
This release is metadata and compatibility only — routing behavior is unchanged from 1.0.1 (same sticky owner/session overrides, static configured routes, cheapest-default, and structured audit stream).
Changed
- Correct catalog listing. The plugin now declares its categories, so it lists under model-routing / cost-optimization instead of "Other." (1.0.0–1.0.1 shipped without category metadata and defaulted to "Other," which made the plugin hard to find by browsing.)
- Canonical source. The package is now linked to its home organization repository, github.com/ToggleLogic/togglelogic-free, replacing the earlier personal-account link.
- Explicit compatibility. Supported OpenClaw: 2026.6.5 – 2026.6.11, validated on 2026.6.11; minimum gateway >=2026.6.5, plugin API >=2026.5.2. (ClawHub no longer infers "latest" compatibility, so supported versions are now declared, not assumed.)
Field note (first-party operational evidence)
Across a four-week window (2026-06-07 to 07-04), two OpenClaw deployments carried the same underlying misconfiguration — an empty static Anthropic model catalog referencing a model absent from OpenClaw's stock built-in catalog. On a production customer deployment running stock OpenClaw without ToggleLogic, the main conversational lane threw 24 "unknown model" failover errors, forced 12 model-fallbacks, silently downgraded every reply to a higher-priced tier (~3x the intended per-token cost), and stamped error banners (with intermittent blank replies) on customer messages — sustained for roughly four weeks. A second deployment with the identical empty catalog but with ToggleLogic present resolved its models cleanly and served traffic normally throughout, with zero forced fallbacks. Same misconfiguration; the deployment with ToggleLogic did not exhibit the failure. (First-party operational evidence, not a lab benchmark.) The full incident analysis is documented internally and is available to licensing partners under NDA.
Roadmap
- Coming in 1.0.3: pin governance — budget-aware handling of user model overrides.