Commit Graph

8 Commits

Author SHA1 Message Date
wizardchen 6d0f09d75f fix(agent): 加固 MCP 工具人审批的跨实例与并发语义
- 跨实例 Resolve 通过 Redis reply channel 接收 owning 实例的 ack
  (ok / not_found / tenant_mismatch / user_mismatch / already_resolved),
  HTTP 不再静默 200,超时退化为 NotFound。
- PendingRequest 增加 UserID,Resolve 校验会话所有者,避免同租户
  其他用户越权批准/篡改参数。
- 新增 ErrAlreadyResolved,timer/ctx 命中后再次 Resolve 返回 400 而
  不是静默成功;waiter.deliver 暴露胜出语义。
- NeedsApproval 默认 fail-close(DB 故障要求审批),可经
  WEKNORA_AGENT_TOOL_APPROVAL_FAIL_OPEN=true 回退。
- Pub/Sub 增加 OriginID 自过滤、退避重连、namespace 后缀,避免自回
  环噪声与多部署串扰。
- 工具执行超时常量统一通过 ToolExecContext.ExecTimeout 注入,
  mcp_tool.go 不再硬编码 60s。
- handler 修正 path 二次解码、错误用 errors.Is、modified_args=null 的
  静默清空、ListMCPToolApprovals 的 404/500 误判。
- repository 用 clause.OnConflict 做原子 upsert,消除并发 500。
- 增加 NotFound / TenantMismatch / UserMismatch / AlreadyResolved /
  Race 等用例。
2026-05-10 22:57:12 +08:00
wizardchen 5510ea8f5a feat(agent): human-in-the-loop approval for MCP tool calls (#1173)
Add an opt-in human approval gate so Agent runs pause before executing
MCP tools that operators flag as dangerous, surface an approval card in
the chat UI, and only resume after the user approves (optionally with
edited args) or rejects.

Backend
- New mcp_tool_approvals table + repo/service to mark per-tool approval
  required (PG migration 000042 + sqlite init).
- approval.Gate coordinates RequestAndWait / Resolve with sync.Once
  delivery, configurable timeout, and Redis Pub/Sub fan-out so multi-
  replica deployments work without sticky sessions.
- MCPTool.Execute integrates the gate; uses a round-level ApprovalCtx
  (without the per-tool 60s timeout) for the wait, and re-derives a
  fresh 60s exec ctx after approval so CallTool keeps a full window.
- New SSE response types (tool_approval_required / _resolved) and
  EventBus events plumb approval state to AgentStreamDisplay.
- REST: list/set per-tool approval flag, resolve pending approval.
- Configurable via agent.tool_approval_timeout_seconds (yaml) or
  WEKNORA_AGENT_TOOL_APPROVAL_TIMEOUT env (accepts seconds or Go
  duration).

Frontend
- MCP settings: per-tool "require approval" switch on the test panel.
- Chat: ToolApprovalCard renders the pause point with editable JSON
  args, validation feedback, mm:ss countdown that turns warning/danger
  near deadline, and a resolved state that retains context.
- i18n strings added for zh-CN / en-US / ko-KR / ru-RU.

Docs
- docs/zh/mcp-approval.md covering behavior, config, API, deployment
  considerations (Redis cross-instance, restart limitations).
2026-05-10 22:57:12 +08:00
wizardchen c12296aa88 feat(observability): instrument agent ReAct loop and tool calls in Langfuse
The existing Langfuse integration covered Chat / Embedding / Rerank / VLM /
ASR generations plus the HTTP + asynq spans, but the agent's own execution
tree was invisible: tool calls never appeared, multi-round ReAct iterations
were flat under the HTTP trace, and there was no single node representing
"one agent run".

This change adds three levels of agent-side spans:

  - agent.execute       — wraps AgentEngine.Execute, records query preview,
                           knowledge bases, allowed tools, final-answer length
                           and totals on finish.
  - agent.round.<N>     — wraps each ReAct iteration; records finish_reason,
                           tool-call count, token usage and duration.
  - agent.tool.<name>   — wraps each tool invocation; records arguments,
                           success, duration, output preview (rune-safe, 4KB
                           cap), error, data keys and image count.

To keep the loop's many exit paths (natural stop, stuck loop, empty-content
retry, final_answer, context cancellation) span-safe, the iteration body was
extracted into runReActIteration with a single defer span.Finish() and an
iterOutcome sentinel driving the outer loop. database_query arguments are
redacted (keys only) to avoid leaking raw SQL into the observability
backend, mirroring the existing UI hint policy.

Adds unit tests for the new helpers (truncateForLangfuse, argKeys, dataKeys,
finishToolSpan nil-safety, iterOutcome.String).
2026-04-24 19:58:08 +08:00
Rene Zander 309fc7b42b feat: support parallel tool calling
When ParallelToolCalls is enabled in AgentConfig and the LLM returns
2+ tool calls, execute them concurrently via errgroup. Results are
collected in original order for deterministic observe-phase behavior.

Sequential execution remains the default. Parallel mode is opt-in
per agent configuration to avoid side effects with stateful tools.

Refactors executeToolCalls into composable helpers (runToolCall,
executeSingleToolCall, executeToolCallsParallel) for testability.

Fixes #707
2026-03-30 10:41:08 +08:00
wizardchen 0e01a8ca64 feat(server): enhance server startup and shutdown process
- Added support for SO_REUSEPORT in the listenWithRetry function to improve port binding during hot-reloads.
- Implemented graceful shutdown by closing the listener immediately upon receiving a shutdown signal, allowing for quicker port release.
- Updated logging to provide clearer feedback during server shutdown and error handling.
2026-03-25 22:08:29 +08:00
wizardchen 08ac3937ac fix(agent): update tool hint comment for clarity in UI progress display 2026-03-25 22:08:29 +08:00
wizardchen e138e0c81a refactor(agent): remove query display from DatabaseQuery component and update tool result structure
- Removed the query display section from the DatabaseQuery component to streamline the UI.
- Updated the DatabaseQueryData interface to eliminate the query field, reflecting changes in tool result handling.
- Enhanced the DatabaseQuery tool execution to focus on returning structured results without exposing raw SQL queries.
- Added a new utility to strip <think> blocks from LLM outputs, improving content clarity and user experience.
2026-03-25 22:08:29 +08:00
wizardchen 637376428e feat(agent): implement core agent functionalities and context management
- Introduced the AgentEngine with methods for executing tool calls, managing context windows, and handling final answer generation.
- Added functionality to stream LLM responses and tool results through the EventBus, enhancing real-time interaction.
- Implemented context management to consolidate messages and prevent exceeding token limits, ensuring efficient use of context.
- Enhanced error handling and logging for tool execution and response analysis, improving robustness and traceability.
- Added unit tests to validate the new functionalities and ensure reliability in agent operations.
2026-03-25 22:08:29 +08:00