v0.122.6
Agenta-AI/agentav0.122.6Oct 15, 2026by mmabrouk
Upgrading?
AI Summary
This release introduces significant improvements including support for the new Gemini 4 Argon model, OpenTelemetry server tracing, and better handling of slow model providers. It also adds frontend capabilities to create agents from scratch or GitHub repos and enhances observability with Google Cloud JSON logging.
Key Highlights
- Added Gemini 4 Argon model support
- Implemented OpenTelemetry request tracing
- Improved handling for slow model providers (timeouts, retries)
New Features
- Create agents from scratch, upload, or GitHub repo
- Add Gemini 4 Argon and price included models
- Server request tracing with OpenTelemetry
- JSON logs for Google Cloud Logging
Full Release Notes
## What's Changed * fix(api): apply the starter-credits email policy to the signup grant and the daily credits by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7554 * fix(api): return session records in the order the runner sent them by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7557 * fix(frontend): adopt the session's own commits and settle replayed tool steps by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7555 * test(api): expect the done record in session replay by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7559 * ci(api): run the OSS integration tests against a real Postgres by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7560 * docs(changelog): Point your agent to a file with @ by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7264 * feat(frontend): create agents from scratch, an upload, or a GitHub repo by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7558 * [fix] Carry Gemini thought signatures in the gateway, not the runner by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7565 * [7522] fix(api): cap the mount text read at 4 MiB and answer 413 above it by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7567 * [7522] fix(frontend): check chat file mentions by folder listing, not a content read by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7568 * [feat] JSON logs for Google Cloud Logging by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7571 * fix(hosting): leave web-mobile out of a --no-web run by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7574 * docs(hosting): explain how to set AGENTA_SERVICES_INTERNAL_KEY by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7576 * fix(gateway): name the model endpoint and quote its error when a model call fails by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7566 * [chore] Remove New Relic by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7569 * fix(sdk): never let a JSON log line raise by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7578 * fix(api): end a dropped LLM gateway stream with an error event, not a 502 by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7575 * feat(api): add Gemini 4 Argon and price included models at 1.2x Google's price by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7570 * fix(chart): optional pod priority and spare capacity placeholders by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7581 * fix(api): handle slow model providers like LiteLLM (timeouts, retries, 408) by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7573 * chore(chart): format the priority and spare capacity test by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7584 * test(sdk): fix CodeQL alert in the JSON logging test by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7590 * [feat] Server request tracing with OpenTelemetry by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7572 * test(api): expect 408 on a gateway timeout and use a valid Slack App ID by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7593 * fix(sdk): template loads fail when the agent uses the built-in model by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7592 * fix(hosting): put the EE gh web service behind the with-web profile by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7596 * test(gate): secret_opaque checks for no real key in the sandbox by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7598 * fix(observability): structure gunicorn, startup and warning logs and name services spans by route by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7594 * fix(api): hide a provider's masked copy of a key in model errors by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7597 * fix(api): report only the spendable credits above the wallet floor by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7599 * fix(runner): send Pi calls on a Gemini provider key to chat completions by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7600 * fix(sdk): show the Gemini 4 Argon display name in the model list by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7602 * fix(services): JSON gunicorn logs in the ee image, and route span names behind the /services prefix by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/7603 * [release] v0.122.6 by @github-actions[bot] in https://github.com/Agenta-AI/agenta/pull/7556 **Full Changelog**: https://github.com/Agenta-AI/agenta/compare/v0.122.5...v0.122.6
Related Repositories
The AI that really does things. Any OS. Any Platform. The lobster way. 🦞
Interactive roadmaps, guides and other educational content to help developers grow in their careers.
Reverse engineer anything with agents, from app behavior down to native binaries.
The open source coding agent.
Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.