v2.0.0.post3

What's Changed

• fix(mcp): migrate ${VAR} headers to env credential refs during driver migration by @xiaoming-qxm in #6091
• refactor(ci): harden desktop workflows and drop legacy verify dead code by @yutai78786 in #6110
• fix(memory): restrict automatic memory to external user queries by @jinliyl in #6120
• feat(plugin/cloudpaw): migrate to QwenPaw 2.0 and bump version to 0.0.4 by @Saint-Yin in #6099
• fix(governance): clear stale OFF-mode sandbox state by @niceIrene in #6122
• fix(runtime): honor per-request model override by @RerankerGuo in #5731
• fix: allow auto_memory_interval to be set to 0 to disable auto memory by @zhaozhuang521 in #6135
• fix(mcp): make legacy MCP migration one-shot via schema watermark (#6130) by @xiaoming-qxm in #6133
• fix(console): require auto_memory_interval as int >= 0, disallow empty by @zhaozhuang521 in #6142
• ci: pass Supabase config to website build by @Osier-Yi in #6143
• fix(utils): add errors='replace' to _run_command for GBK compatibility by @liu6yi1 in #6140
• feat(website): add blog view/like counts and switch GA to QwenPaw property by @yuluo1007 in #6147
• feat(memory): enhance ReMe configuration and indexing safeguards by @jinliyl in #6153
• fix(agents): surface malformed tool-call input to model by @wananing in #5761
• fix(desktop): keep Tauri updates on OSS only by @jinglinpeng in #6071
• fix(desktop): prevent WKWebView from pinning stale console frontend by @zhaozhuang521 in #6107
• fix(chat): preserve whitespace in streaming thinking/text deltas by @zhaozhuang521 in #6166
• fix: images stripped for multimodal models when model info lookup fails or builtin annotation is wrong by @qbc2016 in #6154
• feat(website): add Community Benefits nav and improve blog tags/meta cards by @yuluo1007 in #6164
• fix(website): update architecture overview by @cuiyuebing in #6172
• feat(crons): add silent agent delivery by @AaronZ345 in #6080
• fix(scroll): harden context limits and recoverable tool-result compaction by @niceIrene in #6123
• chore(deps): bumping version of agentscope to 2.0.4.post1 by @qbc2016 in #6179
• fix(loop): fine-tune doom loop thresholds, fix frontend after calc & align tests with #5761 by @rayrayraykk in #6184
• fix(mcp): unblock workspace startup after connection timeout by @jinglinpeng in #6174
• fix(chat): refresh updated_at on user messages and invalidate stale m… by @zhaozhuang521 in #6180
• fix(channels): bound unbounded state and track fire-and-forget tasks to prevent memory leaks by @hongxicheng in #6168
• fix(memory): add dream schedule toggle by @ployts in #6171
• test(ci): run console vitest with coverage in nightly full sweep by @yutai78786 in #6194
• test(e2e): adapt selectors for v2.0.0 UI redesigns by @yutai78786 in #6185
• fix(model_factory): resolve file:// URIs in DataBlock to local paths before formatting by @qbc2016 in #6191
• fix(skill): skill-hub related memory leakage by @Leirunlin in #6208
• fix(utils): drop redundant nvidia-smi probe in get_vram_size_gb by @Yigtwxx in #6204
• fix: sanitize fallback namespace in MCP tool name to comply with OpenAI constraints by @qbc2016 in #6209
• fix(cli): conditionally elevate UAC on Windows so VBS headless launchers stay windowless by @wangfei010313 in #6127
• ci: add CodeQL 2-shard security scan (dry-run) + Dependabot by @hanson-hex in #6027
• fix(memory): bound summary task history by @jinliyl in #6206
• chore: bump the version to 2.0.0post3 by @cuiyuebing in #6211
• feat(console): add history_retention_days config to context compact p… by @zhaozhuang521 in #6214
• fix(website): switch GA tag ID and add Jul 17 developer day session by @yuluo1007 in #6212
• test(integration): Sprint 4.2 — access-control / fork / coding-mode / agent-status by @yutai78786 in #6213
• feat(governance): add rules by @cuiyuebing in #6215
• fix: expand regex shorthands in tool schema patterns for GBNF compatibility by @qbc2016 in #6216
• fix(browser): add MAX_WAITTIME for browser use by @x1n95c in #6170
• fix: treat unprobed multimodal as fail-open to prevent image stripping by @qbc2016 in #6217
• fix(runtime): pass model_slot_override from HTTP request to model factory by @cuiyuebing in #6218
• fix(token_usage): don't persist an unseeded cache on shutdown by @Yigtwxx in #6220
• Refactor channel base by @yuanxs21 in #6159
• feat: bound multi-agent startup and improve readiness UX by @rayrayraykk in #6198

New Contributors

• @liu6yi1 made their first contribution in #6140
• @AaronZ345 made their first contribution in #6080
• @Yigtwxx made their first contribution in #6204

Full Changelog: v2.0.0.post2...v2.0.0.post3