vai generate

Generate production-ready code snippets for integrating Voyage AI embeddings and MongoDB Atlas Vector Search into your application.

Synopsis

vai generate <component> [options]

Description

vai generate creates code files for specific components of a RAG integration. It reads your .vai.json project config to populate the generated code with your actual model, database, collection, and field names.

Components

Component	Description
client	Voyage AI client initialization
connection	MongoDB connection setup
retrieval	Vector search retrieval logic
ingest	Document ingestion pipeline
search-api	Search API endpoint

Targets

Target	Framework	Output
vanilla	Plain Node.js	lib/*.js files
nextjs	Next.js (App Router)	Routes, pages, lib files
python	Python (Flask)	Python modules

Options

Flag	Description	Default
<component>	Component to generate (required)	—
-t, --target <target>	Target framework	Auto-detected or vanilla
-o, --output <path>	Output file path	Default per component
-m, --model <model>	Override model in generated code	From .vai.json
--db <database>	Override database name	From .vai.json
--collection <name>	Override collection	From .vai.json
--field <name>	Override embedding field	From .vai.json
--no-rerank	Omit reranking from generated code	—
--dry-run	Show generated code without writing	—
--json	Output as JSON	—
-q, --quiet	Suppress non-essential output	—

Examples

Generate a retrieval module

vai generate retrieval

Generate for Next.js

vai generate search-api --target nextjs

Generate Python client

vai generate client --target python

Preview without writing

vai generate ingest --dry-run

Tips

• Auto-detection checks for next.config.js, requirements.txt, etc. to pick the right target.
• Run vai init first so generated code uses your project's actual settings.
• For complete starter projects (multiple files + package.json), use vai scaffold instead.

Related Commands

• vai scaffold — Create complete starter projects
• vai init — Initialize project config