Skip to content

v2.0.8

Latest
Latest

Choose a tag to compare

View all tags
@DavdGao DavdGao released this 08 Sep 13:12
v2.0.8
c7d8e31
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Highlight

  • Realtime Voice Agent: Introduced a realtime voice agent with DashScope backends and local transport support.
  • A2A Protocol Support: Added A2AAgent class to the agent module, enabling agent-to-agent communication.
  • New Pipeline Module: Added a new pipeline module with a GoalPipeline class for goal-driven agent workflows.
  • Agent-Driven Context Compression: Introduced a context compression tool that allows agents to manage their own context window.
  • LLM Rerank for RAG: Added LLM-based reranking support to improve RAG retrieval quality.
  • Volcengine Ark Model: Added support for Volcengine Ark chat model.
  • MCP Enhancements: Support runtime MCP headers in Docker workspaces and improved MCP connection cleanup.
  • Web UI Polish: Chat experience improvements including session auto-naming, query caching, and better dark mode support.

What's Changed

  • fix(formatter): derive audio format from media type by @ump45nose in #2301
  • fix(gemini): exclude tool-execution tokens from output_tokens by @iluv7 in #2406
  • feat(formatter): use native multimodal tool outputs by @fancyboi999 in #2389
  • docs(examples): fix broken relative link to _types.py in mem0 README by @feizhuzheng in #2438
  • feat(pipeline): add new pipeline module and a GoalPipeline class by @DavdGao in #2428
  • fix(webui): isolate sessions when switching agents by @iluv7 in #2429
  • chore: bump version to 2.0.7.post1 by @DavdGao in #2452
  • feat(agent): generate a final text summary after reaching max iterations by @qbc2016 in #2443
  • fix(tracing): report interrupted chat responses by @fengting124 in #2450
  • fix(model): preserve encrypted reasoning items in OpenAI Responses history by @qbc2016 in #2426
  • fix(gemini): sanitize nullable type arrays by @helloJamest in #2437
  • fix(webui): keep channel toggle inside card by @iluv7 in #2466
  • fix(webui): follow system dark mode so sidebar buttons stay visible (#2467) by @PiedPiper911 in #2468
  • fix(state): preserve LRU cache under concurrent hits by @fengting124 in #2457
  • fix(xai): include reasoning tokens in output usage by @iluv7 in #2461
  • ci(monitor): stop watching review events that can never access secrets by @DavdGao in #2472
  • fix(formatter): derive multimodal tool-result identifier from the block's own stable id by @AmirF194 in #2165
  • fix(example): disable uvicorn hot reload on Windows by @dvd233 in #2434
  • fix(tool): make read cache backend-aware so Read->Edit works in workspaces by @nolanchic in #2092
  • fix(agent): hint after repeated tool errors by @he-yufeng in #1816
  • fix(app): drain the wake-up queue atomically by @DavdGao in #2476
  • fix(scheduler): let one node own the timers by @DavdGao in #2477
  • feat(rag): add LLM rerank support by @BreezyB1n in #1975
  • fix(scheduler): validate cron before persistence by @redSun64 in #2442
  • feat(agent): add agent-driven context compression tool by @jinliyl in #2143
  • feat(channel): bind channel credentials interactively by @DavdGao in #2484
  • fix(channel): show a channel that is starting as connecting, not stopped by @DavdGao in #2487
  • feat(a2a): add A2AAgent class in the agent module by @YingchaoX in #2142
  • docs(middleware): fix reply budget example class name by @zjncs in #2493
  • fix(tool): apply schema-guided repair to tool arguments by @iluv7 in #2496
  • fix: correct misleading docstring in src/agentscope/init.py by @sunrong1 in #2502
  • fix(rag): normalize distance scores so higher means more relevant by @lyh7c00 in #2486
  • fix(agent): record compression usage in context by @fengting124 in #2433
  • feat(web_ui): chat polish, session auto-naming, and a query cache by @DavdGao in #2503
  • fix(webui): auto-open task panel on first stateUpdated per session by @aniruddhaadak80 in #2431
  • fix(mcp): clean up cancelled MCP connections by @014-code in #2499
  • chore(release): bump version to 2.0.8 and add the A2A news item by @DavdGao in #2505
  • fix: repair agent-facing prompt template text defects by @simpleqt in #2513
  • docs: replace full-width colons with ASCII colons in CONTRIBUTING_zh.md URLs by @simpleqt in #2511
  • fix(app): clear inbox consumer on session delete by @CoralGarden52 in #2519
  • feat(mcp): support runtime MCP headers in Docker workspaces by @qbc2016 in #2456
  • fix: block sending by Enter while a reply is in progress by @DavdGao in #2529
  • fix(rag): escape Excel Markdown table cells by @MaxFreedomPollard in #2528
  • fix(deps): pin ripgrep below 15.x to restore Windows dev installs by @Shxiao101 in #2526
  • refactor(app): a session's origin is one tagged union by @DavdGao in #2536
  • fix(app): support cross-owner agent invites by @logicwu0 in #2326
  • test(mcp): follow SDK transport defaults by @iluv7 in #2542
  • feat(model): add Volcengine Ark chat model support by @iluv7 in #2532
  • feat(realtime): realtime voice agent with DashScope backends and local transport by @DavdGao in #2547
  • docs(zh): link zh news entries to the zh docs locale by @ConcaSun in #2546
  • fix(pipeline): honor verifier context reset by @yaodong-shen in #2544
  • fix(web-ui): keep running MCP tool calls visible by @yang0228 in #2490
  • fix(state): skip oversized read-cache entries by @wuwenbo0626 in #2534
  • refactor(realtime): align RealtimeAgent naming with Agent by @DavdGao in #2550
  • fix(task): hide completed prerequisites from blockedBy by @wuwenbo0626 in #2541
  • fix(rag): preserve blank lines between paragraphs in WordParser by @liugy789 in #2538

New Contributors

Full Changelog: v2.0.7...v2.0.8

Contributors

YingchaoX, yang0228, and 30 other contributors
Assets 2
Loading