Shipped · Observability

What future-agi/future-agi shipped

The public repository of Futureagi · github.com/future-agi/future-agi

Written by FoxPlug from the repository's own releases; not affiliated with Futureagi. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/future-agi/future-agi; Futureagi did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.

Get a weekly update like this for your product, free

Or Use it as a GitHub Action

See the update FoxPlug writes for future-agi, free

Week of November 2, 2026

What shipped

Why it matters

This week added the last parts of the fix my agent flow: a fix that protects cases which must still pass, and resuming an interrupted run without repeating calls that already finished. Gateway fixes made search pricing honest and stopped keys from listing models they cannot call, and dataset writes now reject malformed rows instead of saving blanks. Two releases, 1.57.0 and 1.58.0, went out during the week.

Changelog entry

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

Post for X

This week: resume an interrupted run without repeating finished calls, a fix that protects cases which must still pass, search priced per query, keys no longer listing models they cannot call, and dataset rows that can no longer save blank. v1.57.0 and v1.58.0 shipped.

Post for LinkedIn

This week's work on the project focused on the fix my agent flow and on the gateway. An interrupted run can now be resumed without repeating calls that already finished, on the same run ID and versions. A fix can be given cases that must still pass, and a fix for a failing run can end with a tested new agent version and a pull request for GitHub agents. On the gateway, search requests are priced per query, model listing hides global models from keys that cannot call them, and pricing covers models the existing catalogue missed. Elsewhere, dataset writes reject malformed rows instead of saving blank cells, continuous eval task failures post a Slack alert, and every backend release now flows into the SDK. Releases 1.57.0 and 1.58.0 went out.

Week of October 12, 2026

What shipped

Why it matters

This week fixed error-feed grouping on large projects and closed a tenant isolation gap in SAML login, both of which affected real usage. Releases 1.50.0 and 1.51.0 went out with database migrations, and the dev-to-main syncs kept the hotfix history intact.

Changelog entry

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

Post for X

This week: error-feed grouping works again on large projects, SAML login is tenant-isolated, and 1.51.0 shipped with migrations. In-app tools now share the generated catalog contracts.

Post for LinkedIn

This week the error feed started grouping findings again on projects with many indexed entries, so new findings become Feed cards instead of failing before they are claimed. SAML login gained tenant isolation and tighter session boundaries with new end-to-end tests. Version 1.51.0 shipped with database migrations, and main was merged back into dev after the grouping hotfix. In-app tools now consume the generated MCP catalog, so chat and external clients share the same tool contracts, and the run calls API reports scoring status per call and eval.

Week of October 5, 2026

What shipped

Why it matters

This week the Error Feed stopped losing work: failed investigations and grouping jobs are retried, retired issue URLs redirect, and findings can reach the Feed without a failed requirement. Release v1.48.0 carries the database migrations behind those grouping changes, so the backend image needs manage.py migrate during rollout. The environments API also picked up re-grade and eval-edit routes so finished runs can be acted on without the legacy routes.

Changelog entry

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

Post for X

V1.48.0 ships four tracer migrations for Error Feed grouping. Failed investigations and grouping jobs now retry with bounded backoff instead of ending as final. The environments API can re-grade evals on a finished run and edit one eval, and added evals show as columns.

Post for LinkedIn

This week's work centers on the Error Feed. Failed trace investigations and grouping jobs are no longer final: transient failures retry with bounded backoff, exhausted or permanent ones are parked, dead leases are reconciled during claim polling, and the failure reason now reaches the report. Group IDs survive merges, retired issue URLs redirect to the survivor or the largest split child, and a finding can reach the Feed without a failed requirement. Release v1.48.0 carries four tracer migrations for grouping, three adding columns and one moving data, applied with manage.py migrate when rolling out the backend image. The environments API also gained routes to re-grade evals on a finished run and edit a single eval, and evals added after a run now appear as columns.

Week of September 28, 2026

What shipped

Why it matters

This week shipped improvements to simulation analytics and performance alongside environment configuration flexibility. Several production hotfixes restored e2e testing and the Observe calls list after recent deployments, ensuring the project remains stable for users running large scenario suites.

Changelog entry

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

Post for X

V1.46.0: analytics prioritization, smarter run queries, and environment-level configuration for execution limits and artifact budgets.

Post for LinkedIn

This week's v1.46.0 adds run analytics that separate agent quality from execution health, cuts call detail load time by optimizing run job lookups, and makes execution limits and artifact budgets configurable per environment. We also restored e2e testing and fixed production call lists after recent changes.

Week of September 21, 2026

What shipped

Why it matters

Security fixes prevent leaking API keys from self-hosted installs and remove unintended anonymous bucket access. Environment polling and stop functionality improve the v3 Environments experience. Fixes to CI and build reproducibility help contributors and self-hosters.

Changelog entry

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

Post for X

V1.41.1 ships security fixes for self-hosted installs and improves v3 Environments with live run control and auto-polling.

Post for LinkedIn

V1.41.1 is out. We fixed critical security issues in self-hosted installs—API keys no longer leak to cloud, and anonymous MinIO access is removed. v3 Environments now let you stop live runs from the UI and the list auto-polls during builds. Better contributor experience with improved CI.

Week of September 14, 2026

What shipped

Why it matters

This week shipped critical stability fixes for eval-tasks that were causing excessive cloud storage egress in production. Multiple releases addressed filtering issues on dashboards and preserved evaluation configuration across saves. The project continues to refine alert thresholds and evaluation handling.

Changelog entry

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

Post for X

V1.38.4 is out with fixes for eval-task requeue storms that were costing ~700 GB/day in S3 egress. Also includes dashboard filter fixes and label improvements.

Post for LinkedIn

This week we shipped v1.38.4 with critical fixes for continuous eval-tasks that were re-running converged failures on every poll cycle, causing ~700 GB/day of unnecessary S3 egress in production. We also fixed dashboard filters not applying to eval metric charts and improved the harness UI labeling. Multiple hotfixes ensure stability across the evaluation pipeline.

Weeks with too little public activity are left out rather than filled in. Last updated 2026-11-10.

Is this your repo? Ask us to remove this page.