`ai_chunk_merge`

Merge small text chunks into token-budget groups.

Parameters

source_keystringrequired: Context key holding array of chunk strings
output_keystringdefault "merged": Prefix for output context keys
chunk_sizenumberdefault 512: Target token count per merged chunk

Context Output

Key	Type	Description
`{output_key}`	array	Array of merged chunk strings
`{output_key}_count`	number	Number of merged chunks
`{output_key}_success`	boolean	`true` on success

Algorithm

Uses greedy merging with whitespace-based token counting:

Count tokens per input chunk (whitespace split)
Combine consecutive chunks while total tokens fit within chunk_size
When the next chunk would exceed the budget, start a new group
Groups are joined with "\n\n"

Examples

Merge after chunking

flow:step("chunk", nodes.ai_chunk({
    mode = "split",
    source_key = "document",
    output_key = "parts",
    delimiters = ".?!"
}))

flow:step("merge", nodes.ai_chunk_merge({
    source_key = "parts",
    output_key = "merged",
    chunk_size = 256
})):depends_on("chunk")