local-vectorize

Chunk supported local text and maintain an on-disk retrieval index.

When to use it

Use this node to prepare workflow-specific retrieval. It is independent of Second Brain's folder index and does not enable Second Brain automatically.

Example

Rattish: 1

Workflow:
  name: Build a local index

Node step:
  type: local-vectorize
  source-path: notes
  index-path: indexes/notes.json
  glob: "**/*.md"
  mode: incremental
  finish: pass

Create the source folder and text files before preflight. Incremental mode is the default; full rebuilds and validate checks index currency. The default chunk size is 1,200 with 120 overlap. Keep overlap below the selected size.

The built-in strategy pair is hash_token_v1 and cosine_v1. This is local retrieval with deterministic token hashing, not a claim that a remote embedding model has read the folder. Match the same pair in local-search.

Output distinguishes a current, stale, or updated index and includes file/chunk counts. A query needs an existing readable index, so build the index deliberately before running a separate search workflow.

Configuration

Configuration fields use hyphens. These are the node-specific fields; shared execution fields apply separately.

Field	Type	Required / default	Meaning
source-path	string	Required	Source resource path.
index-path	string	Required	On-disk local retrieval index path.
glob	string	"**/*"	Source-file pattern.
recursive	boolean	true	Permit recursive operation.
chunk-size	integer (min 1)	1200	Chunk size used by the indexer.
chunk-overlap	integer (min 0)	120	Overlap between text chunks.
encoding	string	"utf-8"	Text encoding.
mode	incremental, full, validate	"incremental"	Index update or validation mode.
embedding-strategy	string	"hash_token_v1"	Installed text-to-vector strategy ID.
search-strategy	string	"cosine_v1"	Installed ranking strategy ID.

Inputs and output

This node has no declared input ports. Configure its authored fields directly.

Successful output fields use these exact selector names:

Field	Value type
source_path	string
index_path	string
mode	string
status	current, stale, updated
current	boolean
file_count	integer
chunk_count	integer

For example, refer to node.step.output.source_path from a compatible downstream binding.

Preflight and limits

The source must be readable, the index destination writable, the encoding supported, and the strategy pair available. Runtime scanning and reads have resource limits.

Use Raticode rattish check and Raticode rattish preflight before choosing to run the example. The installed contract is available through Raticode rattish docs --format json.

Continue with Node reference, Routes and joins, or Runs and approvals.