feat: update merge register & with eager execution (#1333)

Author: meguminnnnnnnnn

content/en/docs/eino/core_modules/chain_and_graph_orchestration/orchestration_design_principles.md

@@ -359,6 +359,13 @@ graph.AddLambdaNode("your_node_key", compose.InvokableLambda(func(ctx context.Co
 }), compose.WithOutputKey("your_output_key"))
 ```
 
+You can also register custom merge method to support any type:
+
+```go
+// eino/compose/values_merge.go
+func RegisterValuesMergeFunc[T any](fn func([]T) (T, error))
+```
+
 Workflow can map the output fields of multiple predecessor nodes to different input fields of the successor node. Eino converts the Struct output from each predecessor to a Map before any merge process, still conforming to the above rules.
 
 ### **Streaming Processing**
@@ -464,4 +471,9 @@ When `NodeTriggerMode == AllPredecessor`, the graph executes using the dag engin
 - Support Branch. At runtime, mark the unselected nodes of Branch as skipped, which does not affect the semantics of AllPredecessor.
 - No need for manual SuperStep alignment.
 
+> 💡
+> When NodeTriggerMode is set to AllPredecessor, the node will execute after all predecessors are ready, but it will not execute immediately. Instead, it still follows the SuperStep - running new runnable nodes after a batch of nodes have completed execution.
+>
+> If compose.WithEagerExecution() is passed in Compile, the ready nodes will run immediately.
+
 In summary, the pregel mode is flexible and powerful but comes with additional mental overhead, while the dag mode is clear and simple but limited in application scenarios. In the Eino framework, Chain uses the pregel mode, Workflow uses the dag mode, and Graph supports both; users can choose between pregel and dag.

content/zh/docs/eino/core_modules/chain_and_graph_orchestration/orchestration_design_principles.md

@@ -365,6 +365,13 @@ graph.AddLambdaNode("your_node_key", compose.InvokableLambda(func(ctx context.Co
 }), compose.WithOutputKey("your_output_key"))
 ```
 
+也可以通过注册 Merge 方法来实现任意类型的 merge：
+
+```go
+// eino/compose/values_merge.go
+func RegisterValuesMergeFunc[T any](fn func([]T) (T, error))
+```
+
 Workflow 可以做到多个上游的输出字段映射到下游节点的不同字段。Eino 内部会将上游输出的 Struct 转换为 Map，因此 Merge 依然符合上面的规则。
 
 ### 流式处理
@@ -469,4 +476,9 @@ Eino 支持各种维度的 Call Option 分配方式：
 - 支持 Branch。在运行时，将 Branch 未选中的节点记为已跳过，不影响 AllPredecessor 的语义。
 - 不需要手动对齐 SuperStep。
 
+> 💡
+> 设置 NodeTriggerMode == AllPredecessor 后，节点会在所有前驱就绪后执行，但并不是立即执行，而是依然遵循 SuperStep——在一批节点全部执行完成后再运行新的可运行节点。
+>
+> 如果在 Compile 是传入 compose.WithEagerExecution()，则就绪的节点会立刻运行。
+
 总结起来，pregel 模式灵活强大但有额外的心智负担，dag 模式清晰简单但场景受限。在 Eino 框架中，Chain 是 pregel 模式，Workflow 是 dag 模式，Graph 则都支持，可由用户从 pregel 和 dag 中选择。