Semantic Search & RAG Tool

The Repository Chunker Tool

Break a repository into coherent, size-bounded chunks with metadata so it can be embedded and retrieved accurately — the first step in making a codebase searchable by meaning.

Explore VDF AI Agents
MeaningSemantic, not keyword, recall
GroundedAnswers cite real sources
AssignableTo any knowledge agent
100%On-premise capable
The Retrieval Problem

Your answer exists — retrieval just can’t find it

Private knowledge is only useful if an agent can retrieve exactly the right piece and ground its answer in it. Keyword search misses, hosted RAG can’t touch sensitive data, and ungrounded models make things up.

01

Keyword search misses

The right content is phrased differently than the query.

02

Ungrounded answers

Without retrieval, models invent instead of cite.

03

Scale hides signal

The best chunk is buried among thousands of near-matches.

04

Hosted RAG is off-limits

Sensitive knowledge can’t go to a third-party index.

How the Tool Works

Repository Chunker, without the risk

Capability

What it does

Split a repository into retrieval-ready chunks.

it splits a repository into coherent, size-bounded chunks with metadata, ready for embedding.

Tool
Repository Chunker

Assignable to any agent

ChunkStructuredMetadataRetrieval-ready

How it works

Predictable, inspectable behavior

Designed to be reliable.

chunking respects code structure and attaches path and symbol metadata, so downstream retrieval returns meaningful units rather than arbitrary text slices.

Governed
Policy + Audit

Every call logged

ScopedLoggedGovernedOn-prem

Governance

Private, governed, on-premise

Runs inside your perimeter.

Indexing and retrieval run on-premise or in your sovereign cloud, scoped per tenant and audit-logged, so even sensitive knowledge is searchable and citable without any of it leaving your perimeter.

100%
On-Prem

Per-tenant, logged

On-premRBACAudit logSovereign
Inputs

Parameters

The rag.chunk_repository tool accepts these inputs when an agent calls it. Required inputs are flagged.

Name Type Required Description
repo string Required Repository to chunk.
max_tokens integer
default: 512
Optional Target maximum size per chunk.
overlap integer
default: 64
Optional Token overlap between adjacent chunks.
In depth

How the Repository Chunker tool works in practice

Repository Chunker is a semantic search & rag tool you assign to a VDF AI agent. It splits a repository into coherent, size-bounded chunks with metadata, ready for embedding. Its hallmarks — Chunk, Structured, Metadata — let an agent rely on it as a dependable step in a larger task rather than a brittle one-off script.

Under the hood, chunking respects code structure and attaches path and symbol metadata, so downstream retrieval returns meaningful units rather than arbitrary text slices. It expects repo as required input, so calls are explicit and easy to audit. Every call is scoped to the requesting tenant and written to an audit log, so the capability is safe to run inside a regulated, on-premise environment — the same governance model behind every VDF AI tool.

Teams reach for Repository Chunker when they need to handle index prep, better recall, and pipelines. It rarely works alone — pair it with Embedding Generator, Vector Upsert, and Batch Embed & Upsert to build a complete, governed workflow, then compose those steps into an on-premise VDF AI Network.

Where it pays back

Where Repository Chunker pays back

Index prep

Prepare a codebase for semantic search.

Better recall

Chunk on structure so retrieval returns whole units.

Pipelines

Feed chunks into embedding and upsert steps.

Custom RAG

Build bespoke retrieval over private code.

How VDF AI connects it

Assigned to agents, orchestrated as networks

On VDF AI, an industry’s use cases map to agents, and you assign tools like this one to those agents. Compose multiple agents into a governed, on-premise network.

ROI Snapshot

What changes after you assign it

Faster
To the right knowledge
Cited
Answers traceable to source
Grounded
Less hallucination
100%
Data never leaves your perimeter
FAQ

Questions about the Repository Chunker tool

What is the Repository Chunker tool?

It splits a repository into coherent, size-bounded chunks with metadata, ready for embedding. Assigned to a VDF AI agent, it runs under role-based policy with full audit logging so the capability is safe to use in production.

Why chunk on structure?

Structure-aware chunks keep functions and sections intact, so retrieved context is coherent rather than cut mid-thought.

What comes next?

Embed the chunks and upsert them with the embed and vector-upsert tools to make the repo searchable.

What inputs does the Repository Chunker tool need?

It requires repo, and optionally accepts max_tokens and overlap. Each parameter is validated when an agent calls the tool, and the full call is logged for audit.

Which tools pair well with Repository Chunker?

Repository Chunker is commonly assigned alongside Embedding Generator, Vector Upsert, and Batch Embed & Upsert. On VDF AI you compose several tools and agents into a single governed, on-premise network.

Does it run on-premise?

Yes. Like every VDF AI tool, it can run on-premise or in your sovereign cloud, scoped per user and audit-logged, so your data never leaves your perimeter.

How do agents use it?

You assign the tool to an agent under a role-based policy; the agent calls it as one step in a task, and several agents and tools can be orchestrated together as a governed VDF AI Network.

Put Repository Chunker to work

See the Repository Chunker tool assigned to an agent and orchestrated in a governed, on-premise network.