autogpt-platform-beta-v0.6.54

# 🚀 Release `autogpt-platform-beta-v0.6.54`

**Date:** 9 April 2026

---

## 🔥 What's New?

### New Features

- **#12515** - Replace suggestion pills with themed prompt categories (by @0ubbe)
- **#12571** - Add copy button to user prompt messages in CoPilot (by @Otto-AGPT)
- **#12526** - Spend credits to reset CoPilot daily rate limit (by @majdyz)
- **#12577** - Add shared admin user search + rate-limit modal on spending page (by @majdyz)
- **#12583** - Live timer stats with persisted duration in CoPilot (by @0ubbe)
- **#12512** - Add extended thinking execution mode to OrchestratorBlock (by @majdyz)
- **#12537** - Generic managed credential system with AgentMail auto-provisioning (by @Abhi1992002)
- **#12582** - Add session-level dry_run flag to autopilot sessions (by @majdyz)
- **#12647** - Add generic ask_question copilot tool (by @majdyz)
- **#12650** - Set git committer identity from user's GitHub profile in E2B (by @ntindle)
- **#12569** - Add SQL query block with multi-database support for CoPilot analytics (by @majdyz)
- **#11797** - Update classic AutoGPT for improved day-to-day usability (by @ntindle)
- **#12655** - Set up Codecov coverage reporting across platform and classic (by @ntindle)
- **#12640** - Redesign onboarding wizard with Autopilot-first flow (by @0ubbe)
- **#12622** - Add include_graph option to find_library_agent (by @majdyz)
- **#12581** - Add rate-limit tiering system for CoPilot (by @majdyz)
- **#12649** - Add gitleaks secret scanning to pre-commit hooks (by @majdyz)
- **#12578** - Add create → dry-run → fix loop to agent generation (by @majdyz)
- **#12672** - Add all 12 Z.ai GLM models via OpenRouter (by @Torantulino)
- **#12665** - Add Playwright E2E coverage reporting to Codecov (by @ntindle)
- **#12623** - Add Fast/Thinking mode toggle with full tool parity in CoPilot (by @majdyz)
- **#12667** - Set up React integration testing with Vitest + RTL + MSW (by @0ubbe)
- **#12629** - Add copilot artifact preview panel (by @ntindle)
- **#12328** - Load copilot messages from newest first with cursor-based pagination (by @kcze)
- **#12691** - Add /orchestrate skill — Claude Code agent fleet supervisor with spare worktree lifecycle (by @majdyz)
- **#12696** - Cost tracking for system credentials (by @majdyz)

### Enhancements

- **#12566** - Admin rate limit check and reset with LD-configurable global limits (by @majdyz)
- **#12592** - Normalize agent instructions for Claude and Codex (by @ntindle)
- **#12601** - Clarify ambiguous agent goals using find_block before generation (by @anvyle)
- **#12686** - Refine AutoPilot onboarding — branding, auto-advance, soft cap, polish (by @0ubbe)
- **#12697** - Remove stale feature flags and stabilize share execution (by @0ubbe)
- **#12662** - Enhance pr-address skill with detailed codecov coverage guidance (by @majdyz)
- **#12692** - Improve pr-test skill — inline screenshots, flow captions, and test evaluation (by @majdyz)
- **#12701** - Fix stale-review gate and add pr-test evaluation rules to SKILL.md (by @majdyz)
- **#12704** - Harden agent fleet scripts — idle detection, pagination, fake-resolution guard, parallelism (by @majdyz)

### UI/UX Improvements

- **#12519** - Auto-reconnect copilot chat after device sleep/wake (by @0ubbe)
- **#12557** - Marketplace card description 3 lines + fallback color (by @0ubbe)
- **#12534** - Hide placeholder text when copilot voice recording is active (by @Abhi1992002)
- **#12532** - Fix array field item layout and add FormRenderer stories (by @Abhi1992002)
- **#12564** - Replace unrealistic CoPilot suggestion prompt (by @Otto-AGPT)
- **#12504** - Show all agent outputs instead of only the last one (by @majiayu000)
- **#12428** - AutoPilot notification follow-ups — branding, UX, persistence, and cross-tab sync (by @kcze)
- **#12638** - Allow horizontal scroll for JSON output data in builder (by @Otto-AGPT)

### Bug Fixes

- **#12548** - Prevent graph execution stuck + steer SDK away from bash_exec (by @majdyz)
- **#12560** - Fix prod Sentry errors and reduce on-call alert noise (by @majdyz)
- **#12514** - Add sink input validation to AgentValidator (by @anvyle)
- **#12551** - Move placeholder_values exclusively to AgentDropdownInputBlock (by @majdyz)
- **#12561** - Upgrade Docker Node.js from v21 (EOL) to v22 LTS (by @majdyz)
- **#12565** - Fix prod Sentry errors and reduce on-call alert noise (by @majdyz)
- **#12574** - Preserve thinking blocks during transcript compaction (by @majdyz)
- **#12593** - Raise AIConditionBlock errors instead of swallowing them (by @majdyz)
- **#12579** - Resolve host-scoped credentials for authenticated web requests (by @majdyz)
- **#12555** - Disambiguate duplicate tool names in OrchestratorBlock (by @majdyz)
- **#12546** - Validate email recipients in Gmail blocks before API call (by @Krishnachaitanyakc)
- **#12540** - Add empty choices guard in extract_openai_tool_calls() (by @goingforstudying-ctrl)
- **#12545** - Validate non-empty input in AIConversationBlock before LLM call (by @Krishnachaitanyakc)
- **#12604** - Add tool call circuit breakers and intermediate persistence in CoPilot (by @majdyz)
- **#12580** - Fix dry-run simulation showing INCOMPLETE/error status (by @majdyz)
- **#12606** - Remove LaunchDarkly local storage bootstrapping (by @seer-by-sentry[bot])
- **#12632** - Prevent duplicate block execution from pre-launch arg mismatch (by @majdyz)
- **#12625** - Prompt-too-long retry, compaction churn, model-aware compression, and truncated tool call recovery (by @majdyz)
- **#12628** - Fix copilot credential loading across event loops (by @majdyz)
- **#12642** - Detect prompt-too-long in AssistantMessage content and ResultMessage success subtype (by @majdyz)
- **#12595** - Rename placeholder_values to options on AgentDropdownInputBlock (by @Torantulino)
- **#12646** - Fix tool output file reading between E2B and host (by @majdyz)
- **#12635** - Recognize Agent tool name and route CLI state into workspace (by @majdyz)
- **#12575** - Propagate dry-run mode to special blocks with LLM-powered simulation (by @majdyz)
- **#12656** - Use OpenRouter client for simulator to support non-OpenAI models (by @majdyz)
- **#12663** - Make frontend coverage checks informational (non-blocking) (by @ntindle)
- **#12660** - Prevent duplicate side effects from double-submit and stale-cache race (by @majdyz)
- **#12688** - Skip AI blocks without model property in fix_ai_model_parameter (by @anvyle)
- **#12709** - OrchestratorBlock dry-run credentials + Responses API status field (by @majdyz)
- **#12708** - Suppress Sentry noise from expected 401s in OnboardingProvider (by @majdyz)
- **#12707** - Make User RPC model forward-compatible during rolling deploys (by @majdyz)
- **#12713** - Wrap PlatformCostLog metadata in SafeJson to fix silent DataError (by @majdyz)
- **#12714** - Convert AttributeError to ValueError in _generate_schema (by @majdyz)

### Documentation

- **#12587** - Remove extraneous whitespace in README (by @carsonkahn-external)

### Dependencies & Maintenance

- **#12591** - Add /open-pr and /setup-repo skills (by @ntindle)

---

## 🎉 Thanks to Our Contributors!

A huge shoutout to everyone who contributed to this release. Special thanks to our new contributors:

- @anvyle
- @carsonkahn-external
- @Krishnachaitanyakc
- @goingforstudying-ctrl

As well as our returning contributors:
- @majdyz
- @0ubbe
- @ntindle
- @Abhi1992002
- @Otto-AGPT
- @Torantulino
- @kcze
- @majiayu000
- @seer-by-sentry[bot]

---

## 📥 How to Get This Update

To update to this version, run:

```bash
git pull origin autogpt-platform-beta-v0.6.54
```

Or download it directly from the  
[Releases page](https://github.com/Significant-Gravitas/AutoGPT/releases).

For a complete list of changes, see the [Full Changelog](https://github.com/Significant-Gravitas/AutoGPT/compare/autogpt-platform-beta-v0.6.53...autogpt-platform-beta-v0.6.54).

---

## 📝 **Feedback and Issues**
If you encounter any issues or have suggestions, please [join our Discord](https://discord.gg/autogpt) and let us know!