Shipped · Observability

What FailproofAI/failproofai shipped

The public repository of Befailproof · github.com/FailproofAI/failproofai

Written by FoxPlug from public releases; not affiliated with Befailproof. An automatic summary of the public release, pull request and commit data of github.com/FailproofAI/failproofai. Befailproof did not write it and does not use or endorse FoxPlug. Every line links to the public change it describes.

Get a weekly update like this for your product, free

Or Use it as a GitHub Action

Follow failproofai's weekly shipped digest

Week of September 21, 2026

What shipped

v1.0.8 adds Jev, a classifier that reads tool calls against what you asked for and judges them above regex policies. Release
Two-tier hook evaluator stacks regex policies as a hard floor with optional semantic review on marked policies, off by default. Pull request #833
v1.0.7 stable cut: failproofaid now records human_input events on every harness and ingests OpenClaw 2026.9.6 compressed transcripts. Release
TypeScript SDK added at sdk/typescript, published as @failproofai/sdk with the same 15 events and wire format as the Python SDK. Pull request #830
@failproofai/sdk 0.0.1-beta.0 released: Node and Python agents now write into one pipe with identical telemetry. Release
failproofaid records what a person typed as human_input for every harness, enabling sentiment scoring in downstream dashboards. Pull request #839
fp CLI adds close, archive, and bulk-clear commands for managing issues after agent changes. Pull request #815
TypeScript SDK integration fixtures cleared of 32 Dependabot alerts without framework changes. Pull request #837
fp-cloud-cli 0.0.1 stable released on PyPI as first stable release. Release
WorkerRuntime now allows hosts to install their own managed-source compiler instead of hard-coding Python expressions. Pull request #829

Why it matters

v1.0.8 brings semantic classification of tool calls above regex baselines [0][1], v1.0.7 stable now records what users actually typed for downstream analysis [5][8], and the TypeScript SDK reaches beta so Node and Python agents share one telemetry pipe [11][12].

Changelog entry

v1.0.8: Jev integration for semantic tool call evaluation above regex policies Release
Two-tier hook evaluator: regex floor plus optional semantic tier, off by default Pull request #833
v1.0.7: human_input events recorded on every harness; OpenClaw 2026.9.6 transcript support Release
@failproofai/sdk 0.0.1-beta.0: TypeScript telemetry SDK with 15 events and unified wire format Release
failproofaid records typed input as human_input for sentiment analysis integration Pull request #839
fp CLI: close, archive, and bulk-clear commands for issue management Pull request #815
WorkerRuntime: customizable managed-source compiler instead of hard-coded Python Pull request #829

Example posts FoxPlug drafted from these changes. Not written or posted by the project.

Post for X

v1.0.8 ships Jev: a classifier that reads tool calls against what you asked for, judges them above regex policies. v1.0.7 stable records human input on every harness. TypeScript SDK in beta.

Post for LinkedIn

This week: v1.0.8 adds Jev, a classifier that reads each tool call against what you actually asked for and judges it above regex policies [0]. v1.0.7 stable released—failproofaid now records human_input events on every harness [5]. TypeScript SDK reached 0.0.1-beta.0, so Node and Python agents write identical telemetry into one pipe [11].