knowledge/document/reader/pdf/v0.6.0

session: support update session state (#713)

codeexecutor/container/v0.6.0

session: support update session state (#713)

Release v0.5.0

What's Changed

• examples: show pre-request model switching with runner by @Rememorio in #567
• docs: fix typo and remove dots by @Rememorio in #569
• server/agui: use IsRunnerCompletion instead of IsFinalResponse by @Flash-LHR in #570
• knowledge: add storage layer of pgvector by @hyprh in #568
• {example, model}: update token tailoring formula and refactor example by @Rememorio in #572
• storage: make deafault client builder of pgsql private by @hyprh in #576
• model/openai: do not ignore reasoning content while streaming by @Rememorio in #577
• examples: fix jsonschema comma separator by @liuzengh in #581
• graph: provide external tool example using interrupt/resume by @WineChord in #584
• storage: make default client builders private and simplify registry lookups by @Rememorio in #579
• storage/mysql: implement storage for MySQL client by @Rememorio in #571
• {agent, examples, docs}: implement State for Invocation by @Rememorio in #565
• telemetry: add metrics by @liuzengh in #519
• graph: skip internal state key to avoid map panic by @WineChord in #585
• {model, examples, docs}: support anthropic model by @Flash-LHR in #575
• examples: update example to execute normal tool after external tool by @WineChord in #591
• model: add PromptTokensDetails to observe Prompt/Context caching by @liuzengh in #582
• {storage, memory, examples}: mysql implementation for memory service by @Rememorio in #515
• {evaluation, examples, docs}: support evaluation by @Flash-LHR in #416
• telemetry/appid: add app/agent name registry for report by @WineChord in #597
• feat: support jupyter code execution by @chengjoey in #578
• feat(telemetry): allow overriding OTEL resource service fields and custom attrs via env or options by @minimAluminiumalism in #594
• {server, examples, docs}: add messages snapshot event for agui by @Flash-LHR in #595
• tool: add filtering mechanism by @liuzengh in #590
• server/agui: supports parallel tool calls for agui by @Flash-LHR in #604
• {server/agui, docs}: add agui base url usage document by @Flash-LHR in #601
• graph: provide multi-conditional edges by @WineChord in #596
• {server/agui, examples, docs}: add userID for AG-UI MessagesSnapshotEvent by @Flash-LHR in #602
• agent: add per-run tool filtering support by @bytethm in #580
• session: impl pgsql session by @hyprh in #592
• {processor/content,graph_agent,llm_agent}: support branchFilterMode option and timelineFilterMode option to filter event message by @nomand-zc in #593
• model: Add a variant type: deepseek by @wjf222 in #599
• tool/function: support skip summarization for function tools by @WineChord in #607
• graph: add node callbacks as internal state key and check nil by @WineChord in #609
• {server/agui, docs}: support RunOptionResolver for AG-UI by @Flash-LHR in #612
• flow: lower error log to warn by @WineChord in #610
• tool: print warning log when name or description is empty by @Flash-LHR in #614
• {graph,state_schema}: allow disabling deep copy per key by @nomand-zc in #613
• feat(telemetry): add dual time-to-first-token metrics and stream request attribute by @minimAluminiumalism in #611
• {model, docs, examples}: support model provider by @Flash-LHR in #600
• knowledge: support coditioned filter and mult documents in search tool by @hyprh in #586
• readme: add Star History chart by @Rememorio in #619
• {memory, examples}: postgres implementation for memory service by @Rememorio in #598
• feat(embedder): implement ollama embedder by @chengjoey in #603
• server/agui: add log for agui by @Flash-LHR in #615
• {examples, docs}: add provider example by @Flash-LHR in #618
• event: add JSON marshal/unmarshal to avoid field shadowing by @Flash-LHR in #588
• server/agui: unique user messages based on requestID in AGUI MessagesSnapshot by @Flash-LHR in #616
• memory/mysql: remove UTC conversion for timestamps by @Rememorio in #622
• session: Remove incorrect UTC conversion, alter app_state/user_state to text by @hyprh in #623
• model:Add a variant type: qwen by @pherzheyu in #625
• {session/postgres}: add timeout for async persistence by @nomand-zc in #620
• session/summary: align PostgreSQL session summary logic with the standard by @nomand-zc in #621
• runner: implement Close method for resource cleanup by @Rememorio in #628
• session: add WaitGroup to ensure async workers closing correctly by @Rememorio in #630
• knowledge: support custom index param in pgvector by @hyprh in #627
• memory: fix Delete and Clear tools not having a default creator by @Rememorio in #633
• {doc, model}: support token tailoring for anthropic by @Rememorio in #606
• session: export local session operations to enable users to implement custom session service by @nomand-zc in #632
• Docs: Add the introduction documentation of variant by @wjf222 in #624
• doc(toolset): add documentation for custom toolset naming option in NewMCPToolSet by @liuzengh in #636
• model/openai: remove duplicate chatResponseCallback calls by @Flash-LHR in #635
• {docs, examples}: update runner cleanup instructions by @Rememorio in #639
• docs: keep docs up to date with code by @WineChord in #640
• processor/content: fix user message loss issue in TimelineFilterCurrentInvocation mode by @nomand-zc in #637
• {llmflow, graph}: fix error always nil in AfterModel and AfterAgent callbacks by @bytethm in #641
• skill: implement preview of agent skills by @WineChord in #573
• examples: add HTTP headers example using HTTPBeforeRequest by @bytethm in #574
• knowledge: support remote embedding of tcvector by @hyprh in #605
• add:model qwen enabel thinking by @pherzheyu in #634
• knowledge: enhance markdown chunking by @hyprh in #638
• processor/content: fix ObjectType mismatch by @nomand-zc in #646
• {agent, server}: set response IDs and object types in a2a by @Flash-LHR in #650
• llmagent: set the wrapperChannel's capacity to match the original channel's capacity by @nomand-zc in #649
• server/agui: filter event with empty content by @Flash-LHR in #647
• graph: prevent stale interrupt resume data from being reused by @Flash-LHR in #642
• server/debug: keep graph completion event in streaming mode by @WineChord in #648

New Contributors

• @chengjoey made their first contribution...

Release storage/tcvector/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)

Release storage/redis/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)

storage/postgres/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)

storage/mysql/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)

storage/elasticsearch/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)

session/redis/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)

session /postgres/v0.5.0

session/postgres/v0.5.0

server/debug: keep graph completion event in streaming mode (#648)