consumeUntilInterrupt consumes events from the iterator until an interrupt is found or it's exhausted.

defaultTokenCounter estimates tokens, which treats one token as ~4 characters of text for common English text. github.com/tiktoken-go/tokenizer is hig

extractToolIdentifiers extracts the tool name and call ID from an AgenticMessage that contains a tool result content block. Assumes one tool result pe

handleRunFuncError is the common error handler for buildNoToolsRunFunc and buildReActRunFunc. It handles compose interrupts (both cancel-triggered and

hasToolCalls checks if an assistant message contains tool calls.

isOwnGraph reports whether the callback is being invoked for the React agent's own graph. After the first onGraphStart call records h.ownAddress, sub

isSystemMsg checks if a message has system role.

isToolResultOnlyMsg checks if a message is exclusively a tool result message (no other content besides tool results). For *schema.Message: role == Too

isUserMsg checks if a message has user role (and is not a tool-result message).

nextIdx. get the next idx for the chain. chain key is: node_idx => eg: node_0 => represent the first node of the chain (idx start from 0) if has paral

preprocessADKCheckpoint fixes a gob incompatibility when resuming old ChatModelAgent/DeepAgents checkpoints. Background - ADK checkpoints are gob-enc

setToolCallArguments updates the arguments for a tool call at the given block index.

setToolResultFromOutputParts converts ToolOutputPart slice back to FunctionToolResultContentBlock slice and sets f.Content.

toolResultToOutputParts converts a FunctionToolResult's Content blocks to ToolOutputPart slice.

truncateSkillContent truncates skill content to fit within maxTokens. It keeps the first portion of the content and appends a truncation marker (e.g.

truncateString truncates a string to maxLen characters, adding "..." if truncated

wrapStreamWithCancelMonitoring wraps a stream with cancel monitoring. When immediateChan fires (CancelImmediate or timeout escalation), the output str

AddIndexer adds an indexer node to the parallel. eg. indexer, err := volc_vikingdb.NewIndexer(ctx, &volc_vikingdb.IndexerConfig{ Collection: "my_c

AddLoader adds a loader node to the parallel. eg. loader, err := file.NewLoader(ctx, &file.LoaderConfig{}) p.AddLoader("output_key01", loader)

AddRetriever adds a retriever node to the parallel. eg. retriever, err := vikingdb.NewRetriever(ctx, &vikingdb.RetrieverConfig{}) p.AddRetriever("o

AfterModelRewriteState is called after each model invocation. The input state includes the model's response as the last message. The returned state is

AgenticModel sets the agentic chat model handler for the handler helper, which will be called when the agentic chat model component is executed.

AppendEmbedding add a Embedding node to the chain. e.g. embedder, err := openai.NewEmbedder(ctx, config) if err != nil {...} chain.AppendEmbedding

ChatModelWithTools returns a chat model configured with tool schemas. If a ToolCallingChatModel is provided, it is used directly (and optionally confi

ConcatAgenticMessagesArray concatenates multiple streams of AgenticMessage into a single slice of AgenticMessage.

ConcurrentRetrieveWithCallback concurrently retrieves documents with callback.

ConvAgenticCallbackOutput converts the callback output to the agentic prompt callback output.

ConvertCallbackHandlers converts []host.MultiAgentCallback to callbacks.Handler.

DefaultFinalize is the default TypedFinalizeFunc implementation, providing the same summary post-processing as the middleware does internally: 1. Repl

DeleteRunLocalValue removes a value that was set during the current agent Run() invocation. This function can only be called from within a ChatModelA

Embedding sets the embedding handler for the handler helper, which will be called when the embedding component is executed.