# Direct Corpus Interaction: Agents Need a Terminal, Not Just a Vector DB

Category: Agent Orchestration
Stack: VentureBeat, Databricks DCI, Sandboxed CLI

## Overview
Why the industry is shifting from static embedding queries to active command execution against data repositories.

## The Problem
Vector search treats repositories as passive text chunks, unable to query schemas, inspect metadata, or run SQL transformations.

## Production Rule
Equip data agents with direct query terminals and database clients. Let them inspect catalog tables and test SQL queries iteratively before returning answers.

## Code / Specification
File: `dci_concept.md`

```industry analysis
Vector Retrieval (Passive):
Query -> Top 5 document chunks -> Probabilistic text answer

Direct Corpus Interaction (Active):
Query -> Inspect database schema -> Generate SQL -> Execute query -> Validate output -> Answer with source proof
```