v0.0.6

v0.0.6 releases commit 43e13c10cbe5d7a443c7481157eaf5e0be4e5e7c and includes the changes since v0.0.5.

Highlights

• Console V2 with native agent, evaluation, observability, governance, and administration pages, plus shared announcements and module tutorial videos.
• A solution architect assistant with reviewed deployment proposals, evaluation asset creation, prompt recommendations, and Harness version canaries.
• Skill Lab task generation, evaluation, training, multi-file inputs, artifact inspection, and publication to the Registry.
• Expanded evaluation workflows: online evaluation, custom Lambda evaluators, simulated personas, CloudWatch log datasets, prompt optimization, and paired canary scoring.
• BYOC agent creation, managed session storage, AgentCore Identity integration, inbound JWT, and outbound user authorization.
• Resource creator tracking and workspace isolation, consistent ledger backups, dependency security updates, and fixes for permission overrides and timestamp display.

What's Changed

• feat(evaluation): use native Claude SDK telemetry by @xiehust in #31
• feat(skill-lab): foundation — vendored SkillOpt, exec-worker infra, Bedrock judge by @xiehust in #32
• feat(skill-lab): task sets — entity, CLI-parity validation, console surface by @xiehust in #36
• feat(skill-lab): evaluation jobs — runner, results console, Registry deep link by @xiehust in #37
• feat(skill-lab): training jobs — gated optimization, diff review, publish to Registry by @xiehust in #38
• feat(skill-lab): Codex CLI target backend + judge model widening by @xiehust in #39
• feat(skill-lab): AI taskset generation (taskgen job type) by @xiehust in #40
• feat(skill-lab): unified select styling + built-in demo samples by @xiehust in #41
• feat(skill-lab): agentic judge (host-side, upstream parity) by @xiehust in #42
• feat(evaluation): skill evaluators + third-party/derived evaluators by @xiehust in #43
• fix(frontend): throw on unparseable 200 bodies instead of resolving null by @xiehust in #44
• fix(frontend): API cache poisoning — no-store on backend calls + body snippet in invalid-JSON errors by @xiehust in #45
• feat(frontend): DiffPanes renders a line-aligned split diff by @xiehust in #46
• feat(skill-lab): default judge gpt-5.6-sol + codex-backed agentic judge by @xiehust in #47
• feat(skill-lab): multi-doc training via trainable_files by @xiehust in #48
• fix(skill-lab): resolve uv alias symlinks in the venv interpreter by @xiehust in #49
• fix(workspaces): default-workspace resource merges mirror into the yaml by @xiehust in #50
• feat(skill-lab): binary and multi-file task assets by @xiehust in #51
• feat(skill-lab): accept md/txt/csv as task input files by @xiehust in #52
• feat(skill-lab): taskgen input attachments by @xiehust in #53
• fix(skill-lab): agentic judge routing, readiness, and diagnosability by @xiehust in #54
• fix(evaluation): preflight ground-truth judges and surface batch failure causes by @xiehust in #56
• feat(evaluation): per-agent online evaluation configs (?view=online) by @xiehust in #59
• feat(evaluation): online scores on session detail + Overview online-quality tile (P3a) by @xiehust in #60
• feat(evaluation): insights-mode online configs + recurring reports (P3b) by @xiehust in #61
• feat(optimization): pluggable recommend providers + gepa_lite reflection by @xiehust in #64
• feat(optimization): gepa_lite also optimizes the agent's own tool descriptions by @xiehust in #65
• fix(optimization): truncation-aware reflection retry; regenerate keeps the pinned run by @xiehust in #66
• fix(optimization): accept defaults to the recommended tool descriptions; tolerant reflection JSON by @xiehust in #68
• chore(optimization): default reflection model → Claude Opus 5 by @xiehust in #69
• fix(experiment): remember the recommend pickers per experiment by @xiehust in #70
• fix(optimization): stream the reflection call with a long read timeout by @xiehust in #71
• Catch-all route: unknown URLs render a not-found view inside the shell by @xiehust in #72
• Unreachable backend: health chip reflects reality and list pages show load errors, not empty states by @xiehust in #73
• Stale "Phase" copy and unlabeled form controls by @xiehust in #74
• Small viewports: no horizontal page scroll below 720 px by @xiehust in #75
• AWS ClientErrors on console routes map to 4xx envelopes, never a bare 500 or raw boto text by @xiehust in #76
• Disabled primary actions explain what is missing (shared Btn disabledReason) by @xiehust in #77
• Stale deep links tell the user the resource is gone instead of silently falling back by @xiehust in #78
• zh-CN typography: full-width punctuation inside Chinese copy, consistently by @xiehust in #79
• Evaluation runs can be stopped (StopBatchEvaluation + queued cancel) by @xiehust in #80
• Agent detail lists AWS versions and endpoints (Runtime + Harness) by @xiehust in #81
• Chat can end the AgentCore Runtime session (StopRuntimeSession) by @xiehust in #82
• Memory resources: edit description and event expiry (UpdateMemory) by @xiehust in #83
• Governance gateway detail manages Gateway rate limits by @xiehust in #84
• Cloud datasets: re-sync edits the AWS draft in place; drafts show status, publish and list versions by @xiehust in #85
• Cloud dataset runs can pin a published dataset version by @xiehust in #86
• Governance gateway detail can synchronize dynamic MCP-server targets (SynchronizeGatewayTargets) by @xiehust in #87
• A2A registry records show the live agent card served by the runtime (GetAgentCard) by @xiehust in #88
• Bootstrap memory carries session-summary and episodic (reflection) strategies by @xiehust in #89
• Chat memory rail: list_events accepts the pinned memory_id again by @xiehust in #90
• Governance gateway detail names each target's kind (MCP / HTTP / inference variants) by @xiehust in #91
• Evaluators: create, show and safely edit code-based (Lambda) custom evaluators by @xiehust in #92
• Registry consumer view: what ListDiscoverableRegistryRecords exposes vs the control-plane list by @xiehust in #93
• A2A runtime card and registry record agree on version by @xiehust in #94
• Observability session detail can score a session on demand (data-plane Evaluate) by @xiehust in #95
• Default model and Registry-GA facts corrected in troubleshooting, setup and README (en + zh) by @xiehust in #96
• AGENTS.md re-mirrors CLAUDE.md; both fix the route list and the make dev port claim by @xiehust in #97
• api.zh-CN.md documents the Governance, Memory and Accounts console routes the English page has by @xiehust in #98
• Deploy job-level failures fail the agent honestly and reach the process log by @xiehust in #99
• Teardown removes the bootstrap gateways, credential providers and orphaned per-agent roles it promises to sweep by @xiehust in #100
• architecture.zh-CN.md gains the nine missing sections and the GA Registry row by @xiehust in #101
• Verify gate runs the backend suite in parallel (pytest-xdist): 6.4 min → ~1 min by @xiehust in #102
• Console routes load lazily: entry chunk drops from 1.9 MB to the shell + Overview by @xiehust in #103
• setup.md names the sandbox, codegen, Skill Lab and AgentCore-timeout settings by @xiehust in #104
• Tracked docs stop linking into the ignored .trellis/ tree; smoke-test route and zh memory-strategy facts corrected by @xiehust in #105
• Knowledge Bases documented: architecture section + mapping row, README bullet, api.md console section, troubleshooting entry (en + zh) by @xiehust in #106
• Memory short-term timeline preserves JSON event payloads by @xiehust in #108
• Skill Lab evaluation results show reported target and judge token usage by @xiehust in #109
• Skill Lab artifacts support live inspection and safe Markdown previews by @xiehust in #110
• Skill Lab generated tasks can be edited and selected before import or expansion by @xiehust in #112
• Protected system-managed AWS solution architect Harness preset by @xiehust in #113
• Launchpad assistant with reviewed idempotent Harness deployment proposals by @xiehust in #114
• Architect live-smoke reliability: bounded model output and accurate retention advice by @xiehust in #115
• Correct native Harness reasoning passthrough to raw Converse fields by @xiehust in #116
• Register protected system preset Skills as reusable Registry entries by @xiehust in #118
• Simplify system preset card and reuse the existing Agent editor by @xiehust in #119
• Show the protected Skill name and installed version in the shared editor by @xiehust in #120
• Executable multi-actor and multi-session golden test scenarios by @xiehust in #121
• Reviewed assistant Dataset and custom evaluator asset creation by @xiehust in #122
• Tag the CDK execution role as platform-managed for reviewed evaluator grants by @xiehust in #123
• Handle Lambda initialization revisions and audited recovery of approved evaluator assets by @xiehust in #124
• BYOC creation method, manylinux_2_28 dependency resolution, free-form runtime replies, list-first agent management by @alex-sample-code in #128
• feat(videos): add production BYOC walkthrough by @xiehust in #132
• feat(playground): support capability-aware file attachments by @xiehust in #133
• feat(console): Console V2 — enterprise-SaaS shell and native evaluation module by @xiehust in #136
• feat(console-v2): classic modules in the V2 shell + native agents, online evaluation, experiments and canaries by @xiehust in #137
• fix(console-v2): keep every stage result visible on finished experiments by @xiehust in #138
• feat(console-v2): native Agent 开发 / Agent 运行 modules by @xiehust in #139
• feat(console-v2): native system admin pages by @xiehust in #140
• feat(create-agent): default to native Bedrock with GPT-6 Sol; add GPT-6 + Kimi K3 by @xiehust in #142
• feat(console-v2): workbench on the anchor pattern + announcement board by @xiehust in #141
• feat(studio): default new canvas agents to Bedrock GPT-6 Sol; flow-scoped execution roles by @xiehust in #143
• feat(console-v2): native agent edit; re-publish keeps non-form spec fields by @xiehust in #144
• Add admin-managed video directory aligned with V2 navigation by @xiehust in #145
• Separate V2 and classic tutorial videos by @xiehust in #146
• feat(agents): lock a converted agent's baked prompt and model on edit by @xiehust in #147
• feat(assistant): proposals default to GPT-6 Sol like the Create Agent wizard by @xiehust in #148
• feat(console-v2): AgentOps logo and "智能体Ops平台" product chip by @xiehust in #149
• fix(classic): keep finished experiments' stage results visible; honest memory chip by @xiehust in #150
• feat(frontend): AgentOps tab icon; Skill Lab under Agent 评估; 实验 → A/B实验 by @xiehust in #151
• fix(videos): file Skill Lab videos under Agent evaluation by @xiehust in #152
• fix(evaluation): DeepEval PIILeakage is higher-is-better by @xiehust in #153
• fix(runtime): recover SDK repr-fallback stream events instead of showing them by @xiehust in #154
• fix(evaluation): wait up to 2 h for a batch and re-check runs that gave up by @xiehust in #155
• fix(convert): restore native Harness tools the CLI export drops by @xiehust in #156
• test(evaluation): wrap an over-long line in test_run_recheck by @xiehust in #157
• feat(console-v2): 评估总览 rename + insights mode in 创建评估任务 by @xiehust in #158
• fix(assistant): no blank fishbone bones; eval-asset plans fit AWS limits by @xiehust in #159
• fix(registry): mount cross-region IAM gateways with SigV4 by @xiehust in #160
• feat(assistant): show conversation creator; admins can share read-only by @xiehust in #161
• feat(assistant): shared conversations are collaborative; preparation before proposal by @xiehust in #162
• feat(console): V2 is the default console; sign-in pages wear V2 by @xiehust in #163
• feat(assistant): in-reply proposal validation, change-based revisions; architect preset GPT-6 Astra + shell by @xiehust in #164
• fix(platform): honest chat memory label, record-scoped gateway tools, GPT-6 judge default with fallback by @xiehust in #165
• fix(invoke): qualify user-Gateway tool selectors for remote_mcp naming by @xiehust in #166
• fix(assistant): reserve evaluator names across evaluation plans by @xiehust in #167
• fix(eval): code evaluators count each Gateway call once under its catalog name by @xiehust in #168
• fix(convert): rebuild Gateway MCP clients the CLI export drops for a narrowed Harness by @xiehust in #169
• fix(experiments): A/B traffic reaches a twin's Gateway tools by @xiehust in #170
• fix(convert): rebuild AWS_IAM Gateway clients with a SigV4 transport by @xiehust in #171
• feat(v2-eval): evaluation type column, insights panel, log-stream source and agent-less CloudWatch runs by @xiehust in #173
• feat(v2-eval): recommendations from an evaluation task; experiment tool recs read pinned-run spans by @xiehust in #175
• fix(v2-eval): missing-Harness fallback, every usable target evaluator (推荐 pair), 评估总览 first in the nav by @xiehust in #176
• perf(v2-eval): read Gateway targets in parallel and cache recommendation catalogs by @xiehust in #177
• feat(v2): V2 Select replaces every native dropdown in the V2 console by @xiehust in #178
• feat(v2-assistant): native V2 fishbone in the architect proposal by @xiehust in #180
• feat(v2-assistant): AI-recommendation step + Harness version canary in the architect's next steps by @xiehust in #179
• feat(canary): Harness canaries may complete straight from 50/50 on a treatment win by @xiehust in #181
• feat(v2-rec): tool-description recommendations render as diffs like the prompt; off by default for a Harness by @xiehust in #182
• fix(v2-assistant): step 3 keeps earlier runs' recommendations visible and lets you pick the source run by @xiehust in #184
• fix(v2-eval): result drawer shows Harness eval transcripts and skips the 7d span query by @xiehust in #186
• feat(canary): a Harness canary may skip 90/10 at creation and complete from a 50/50 tie by @xiehust in #187
• feat(v2-rec): GEPA-lite run recommendations, review-and-edit accept, any-run step 3, dataset row copy, architect edit & retry by @xiehust in #192
• feat(agents): default execution budget 180 s → 600 s by @xiehust in #193
• fix(assistant): architect protocol names 600 s as the proposal budget by @xiehust in #194
• feat(agents): a new Managed Harness starts with shell + file operations by @xiehust in #195
• fix(i18n): native tools hints say a new Harness starts with both ticked by @xiehust in #196
• feat(assistant): adversarial tests stay out of AI prompt recs; proposals can mount Code Interpreter by @xiehust in #197
• feat(agents): managed session storage on Harness and Strands agents by @xiehust in #198
• fix(assistant): mark every red-team test adversarial, not only injections by @xiehust in #199
• fix(assistant): architect phrases the untrusted-content rule so AI recs accept the prompt by @xiehust in #200
• fix(evaluation): replay a scenario on a transient upstream error instead of failing the run by @xiehust in #201
• fix(optimization): replay requests outlive the 600 s agent budget by @xiehust in #202
• feat(canary): Harness canary replays datasets paired, question by question by @xiehust in #203
• fix(evaluation): replay a scenario whose Harness loop stopped after a tool step by @xiehust in #204
• feat(agents): default max_iterations 10 → 100 by @xiehust in #205
• fix(evaluation): replay a scenario once when its Harness execution timed out by @xiehust in #206
• feat(evaluation): score budget-stopped scenarios; GPT Harness max_tokens default 65536 by @xiehust in #207
• fix(evaluation): replay a scenario whose Harness stream closed empty by @xiehust in #208
• fix(canary): a budget-stopped side keeps its pair in the paired verdict by @xiehust in #209
• fix(canary): paired scores query must not redefine the score field by @xiehust in #210
• fix(evaluation): Harmfulness and Stereotyping are higher-is-better by @xiehust in #211
• feat(assistant): replay an architect turn once after a transient upstream failure by @xiehust in #212
• fix(v2): collapse superseded recommendation cards behind an earlier toggle by @xiehust in #213
• fix(assistant): replay an architect turn once after a spurious max_tokens stop by @xiehust in #214
• fix(evaluation): next-steps run mean is normalized, weighted and keeps code evaluators by @xiehust in #215
• fix(chat): rename Attachments.tsx to avoid case collision with attachments.ts by @yourlin in #172
• fix(memory): validate memory ownership before pinning, reading back or managing a memory (#55) by @xiehust in #216
• feat(identity): AgentCore Identity for agents - Connections, as_user 3LO, inbound JWT and on-behalf-of by @alex-sample-code in #191
• fix(identity): drop as_user grants when their agent or Connection is deleted by @xiehust in #219
• feat(console): rename 我的连接 → 我的授权 and 连接 → 出站凭证 in the UI by @xiehust in #220
• fix(v2): identity forms use the V2 Select instead of native selects by @xiehust in #221
• feat(v2): per-module demo videos in the page header by @xiehust in #222
• feat(chat): show Gateway Cedar denials as a policy-deny card in Harness Chat by @xiehust in #223
• fix(assistant): evaluation-assets creation rides perm:agents.deploy, not admin by @xiehust in #224
• feat(eval): save a manual revision of a prompt recommendation before accepting by @xiehust in #225
• feat(assistant): a conversation's owner can share it with the workspace by @xiehust in #226
• fix(eval): replay a scenario when InvokeHarness fails a runtime health check by @xiehust in #227
• fix(registry): keep a record description up to AWS's 4096 characters by @xiehust in #228
• feat(canary): choose the canary's evaluators and a primary that decides the verdict by @xiehust in #229
• fix(eval): refuse {actual_tool_trajectory} judges for online evaluation by @xiehust in #230
• docs(samples): store-demo CRM sample and the store-assistant presenter script by @xiehust in #231
• fix(security): remove the vendored standalone Strands Studio app by @xiehust in #232
• docs(samples): store-script screenshots and the HR policy-assistant demo script by @xiehust in #236
• docs(samples): link the published demo videos from the store and HR scripts by @xiehust in #237
• docs(samples): make the HR demo-script HTML a single self-contained file by @xiehust in #238
• docs(samples): presenter script for the investment research assistant demo by @xiehust in #240
• fix(v2): run data pipelines in the background; filter and sort eval results by @xiehust in #239
• feat(v2): admins read every architect conversation; link a pipeline's dataset by @xiehust in #241
• feat(architect): fewer intake turns; a blanket "默认" confirms suggested barriers (skill 1.5.8) by @xiehust in #242
• feat(eval): explain partially failed batches; telemetry-only runs seed recommendations by @xiehust in #243
• fix(eval): telemetry-only partial runs read as info (blue), with a shorter note by @xiehust in #244
• fix(eval): slimmer coverage note in the architect's run list by @xiehust in #245
• feat(v2): dataset editor matches classic — scenario/persona forms + JSON import by @xiehust in #247
• feat(eval): pipeline log source — CloudWatch logs + format rule → dataset items by @xiehust in #248
• feat(v2): 数据处理 运行日志 source — log groups + format step with live preview by @xiehust in #249
• fix(eval): log-source field suggestions cover every format in a mixed group by @xiehust in #250
• feat(eval): a failed simulated persona no longer sinks the run; V2 wizard runs simulated datasets by @xiehust in #251
• feat(eval): actor models — GPT-6 Luna (default) + Haiku 5.5, drop Sonnet 4.6 by @xiehust in #252
• feat(v2): architect canary starts at 50/50 by default; "遥测日志（无需 Agent）" by @xiehust in #253
• fix(assistant): clearing a conversation tears down canary dependencies first (#246) by @xiehust in #254
• Pre-restart in-flight inventory: one read-only command names every background item a restart resumes, fails or strands by @xiehust in #255
• Ledger backups are online-consistent, verified and prunable by @xiehust in #256
• Evaluator catalog says when the account's custom evaluators could not be listed by @xiehust in #257
• Best-effort cleanups swallow only "already gone" and log every other failure by @xiehust in #258
• /v1 API keys record creator and last use; offboarding a member disables their keys by @xiehust in #259
• fix(cleanup): staged BYOC upload delete warns on anything but NoSuchKey by @xiehust in #260
• Every user-managed resource records and shows its creator (创建者) by @xiehust in #261
• Admins can isolate a workspace so members see only their own resources by @xiehust in #262
• Agents record their creator and follow workspace isolation, /v1 keys included by @xiehust in #263
• Patch the four critical Dependabot alerts (PyJWT, anyio, tinypool) by @xiehust in #265
• Patch the high-severity Dependabot alerts; bedrock-agentcore moves to 1.18.* by @xiehust in #266
• V2 keeps permission denials and shows ledger times in the right time zone by @xiehust in #267

New Contributors

• @alex-sample-code made their first contribution in #128

Full Changelog: v0.0.5...v0.0.6