Configuration

Damru — sql2nosql is configured via a single JSON file at the project root: sql2nosql.config.json. Copy from sql2nosql.config.example.json and fill in your values.

Config file reference

Example structure:

{
  "connection": "postgres://<username>:<password>@<host>:<port>/<database>",
  "schema": "public",
  "output": "./output",
  "llm": {
    "enabled": false,
    "apiKey": "<OPENAI_API_KEY>",
    "model": "gpt-4.1-mini"
  },
  "mongodb": {
    "uri": "mongodb://<username>:<password>@<host>:<port>",
    "database": "sql2nosql",
    "collectionPrefix": ""
  },
  "migration": {
    "batchSize": 0,
    "dryRun": false,
    "skipOnError": false,
    "progressEvery": 1000
  }
}

---

Required for analysis

Key	Description
connection	Postgres connection string. Required for yarn analyze.
schema	Postgres schema to introspect (default: public).
output	Directory for generated JSON, HTML, and scripts (default: ./output).

---

MongoDB (for running migrations)

Key	Description
mongodb.uri	MongoDB connection URI. Required when running migration scripts (unless migration.dryRun is true).
mongodb.database	Target database name (default: sql2nosql).
mongodb.collectionPrefix	Optional prefix for collection names (default: "").

---

Migration script options

Used by the generated .migrate.js scripts when you run them.

Key	Description
migration.batchSize	0 = load all rows; >0 = paginate with LIMIT/OFFSET (e.g. 5000) to avoid loading huge tables.
migration.dryRun	true = connect and process rows but do not write to MongoDB (safe to test).
migration.skipOnError	true = log row errors and continue; false = fail on first error.
migration.progressEvery	Log progress every N rows (e.g. 1000); 0 = only final summary.

---

LLM (optional)

The LLM section enables AI-powered schema recommendations during yarn analyze. It is optional and off by default.

Config fields

Key	Description
llm.enabled	Set true to enable LLM recommendations (default: false). You can also pass --llm on the CLI.
llm.apiKey	OpenAI API key. Can be set here, or via --llm-api-key or the OPENAI_API_KEY environment variable.
llm.model	Model name (e.g. gpt-4.1-mini, gpt-4). Default: gpt-4.

Advantages of using LLM

• Embedding recommendations — Suggests which relations to embed vs keep as references, with reasoning (e.g. “embed artist in album for read-heavy workloads”).
• Trade-off explanations — Describes pros and cons (denormalization, indexing, sharding) in plain language.
• Optimized NoSQL layout — Output in output/recommend/ and the view HTML include an LLM-optimized schema and per-table insights you can compare to the deterministic analysis.
• Human-in-the-loop — The tool never migrates data or changes schemas itself; it only proposes. You decide what to apply.

When to enable

• Enable when you want suggestions and rationale for embedding strategy, indexing, or document shape.
• Leave disabled if you only need deterministic SQL → NoSQL mapping and migration script generation (no API key required).

Example with LLM enabled

"llm": {
  "enabled": true,
  "apiKey": "sk-...",
  "model": "gpt-4.1-mini"
}

Then run:

yarn analyze --llm

Results appear in output/recommend/ and in the view pages (e.g. “LLM Optimization Recommendations” and “Additional Insights”).

---

Security

• Do not commit sql2nosql.config.json — it contains credentials (Postgres, MongoDB, and optionally the LLM API key). It is listed in .gitignore.
• Use sql2nosql.config.example.json (no secrets) as the template and document only the shape of the config in docs (as on this page).