Agenta

v0.100.4 Feature

This release adds 3 notable features for engineering teams evaluating rollout.

Published 6d LLM Frameworks

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

agents evaluation llm-as-a-judge llm-evaluation llm-framework llm-monitoring

+9 more

llm-observability llm-platform llm-playground llm-tools llmops observability prompt-engineering prompt-management rag-evaluation

Summary

AI summary

Fixes in frontend, api, and sdk improve trace observability, guardrails, and invocation recording.

Changes in this release

Type	Severity	Summary	CVE
Feature
Feature	Low	Show span parameters in the trace overview UI. Show span parameters in the trace overview UI. Source: llm_adapter@2026-05-28 Confidence: high	—
Feature	Low	Filter trace tree to key spans in the frontend. Filter trace tree to key spans in the frontend. Source: llm_adapter@2026-05-28 Confidence: high	—
Feature	Low	Beautify extraction for non‑chat span values in observability table. Beautify extraction for non‑chat span values in observability table. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix
Bugfix	Medium	Extend guardrails to all git-based entities. Extend guardrails to all git-based entities. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Medium	Map OpenInference tool definitions into real tool objects. Map OpenInference tool definitions into real tool objects. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Medium	Keep tool-call pairing when loading chat into playground. Keep tool-call pairing when loading chat into playground. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Medium	Resolve retrieve‑related stable tie‑break issue for latest revision. Resolve retrieve‑related stable tie‑break issue for latest revision. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Medium	Record invocation parameters on the root span in SDK. Record invocation parameters on the root span in SDK. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Medium	Allow dotted reference slugs in tracing ingest for API and SDK. Allow dotted reference slugs in tracing ingest for API and SDK. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Medium	Stop /workflows/revisions/retrieve from silently returning wrong data. Stop /workflows/revisions/retrieve from silently returning wrong data. Source: llm_adapter@2026-05-28 Confidence: high	—
Bugfix	Low	Hide trace children from raw span data in frontend. Hide trace children from raw span data in frontend. Source: granite4.1:30b@2026-05-28-audit Confidence: low	—
Bugfix	Low	Seed parameter defaults for custom URL workflows in frontend. Seed parameter defaults for custom URL workflows in frontend. Source: granite4.1:30b@2026-05-28-audit Confidence: low	—
Bugfix	Low	Resolve every identifying trace reference when opening playground. Resolve every identifying trace reference when opening playground. Source: granite4.1:30b@2026-05-28-audit Confidence: low	—
Bugfix	Low	Fix flaky acceptance tests (app creation, human eval, prompt registry, members invite). Fix flaky acceptance tests (app creation, human eval, prompt registry, members invite). Source: granite4.1:30b@2026-05-28-audit Confidence: low	—

Full changelog

What's Changed

[fix] Extend guardrails to all git-based entities by @junaway in https://github.com/Agenta-AI/agenta/pull/4422
fix(api): map OpenInference tool definitions into real tool objects by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4397
feat(frontend): show span parameters in the trace overview by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4398
feat(frontend): filter trace tree to key spans by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4404
fix(frontend): keep tool-call pairing when loading chat into playground by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4406
feat(frontend): beautified extraction for non-chat span values in observability table by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4410
[fix] Resolve retrieve-related issue 1-12: stable tie-break for latest-revision by @jp-agenta
[research] Additional fixes required for proper entity retrieval by @junaway in https://github.com/Agenta-AI/agenta/pull/4424
fix(sdk): record invocation parameters on the root span by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4423
fix(api,sdk): allow dotted reference slugs in tracing ingest by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4419
fix(api): stop /workflows/revisions/retrieve from silently returning wrong data by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4418
fix(frontend): hide trace children from raw span data by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4417
[fix] Resolve tests and missing checks in guardrails by @junaway in https://github.com/Agenta-AI/agenta/pull/4448
[release] v0.100.3 by @github-actions[bot] in https://github.com/Agenta-AI/agenta/pull/4451
fix(frontend): seed parameter defaults for custom URL workflows by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4440
fix(frontend): resolve every identifying trace ref when opening playground by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4425
[docs] Update API reference docs by @github-actions[bot] in https://github.com/Agenta-AI/agenta/pull/4421
Trace Observability Batch PR by @mmabrouk in https://github.com/Agenta-AI/agenta/pull/4443
fix(tests): resolve flaky acceptance tests (app creation, human eval, prompt registry, members invite) by @bekossy in https://github.com/Agenta-AI/agenta/pull/4458
[release] v0.100.4 by @github-actions[bot] in https://github.com/Agenta-AI/agenta/pull/4452

Full Changelog: https://github.com/Agenta-AI/agenta/compare/v0.100.3...v0.100.4

View diff on GitHub

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

Share on X Share on Bluesky

Track Agenta

Get notified when new releases ship.

About Agenta

LLMOps platform for prompt management, LLM evaluation, and observability. Build, evaluate, and monitor production-grade LLM applications with collaborative prompt engineering.

Agenta

Summary

Changes in this release

What's Changed

Related context

Related tools

Earlier breaking changes