In LlamaIndex
You want full control of the retrieval, prompt, and query-engine internals.
Comparison
nlqdb vs LlamaIndex
Pick LlamaIndex if you want to assemble your own text-to-SQL query engine — its schema retrieval, prompt, and guardrails — over a database you already run. Pick nlqdb if you want NL→SQL working today as a hosted pipeline you embed, with the SQL shown, writes diff-previewed, and a Postgres provisioned for you.
The same goal, two ways.
> monthly active users for the last 6 months
In your HTML
<nlq-data goal="monthly active users for the last 6 months"></nlq-data>A from-scratch LlamaIndex query engine needs a wired DB, model, prompt, and schema retrieval before this works; nlqdb answers it from one English goal over a Postgres it provisioned, SQL shown.
What's different
Four dimensions that decide it
| Dimension | nlqdb | LlamaIndex | Note |
|---|---|---|---|
| Provisions + owns the database (from English) | LlamaIndex wraps a SQLAlchemy engine over a database you already stood up; it provisions nothing. nlqdb materialises a Postgres from the first English goal. | ||
| NL→SQL without building or tuning the engine yourself | LlamaIndex gives you the query-engine classes; you assemble, prompt, and tune them. nlqdb ships the pipeline — nothing to wire. | ||
| Compiled SQL shown with every answer | The query engine exposes the generated SQL in its metadata, but surfacing it to a user is your wiring. nlqdb returns the compiled SQL with every answer by default. | ||
| Fail-closed read-only SQL validation | The docs warn the query tool runs whatever SQL the model emits; restricted roles and sandboxing are yours to add. nlqdb validates against an allowlist and fails closed. |
Show 4 more rows
| Dimension | nlqdb | LlamaIndex | Note |
|---|---|---|---|
| Destructive-op diff preview before apply | LlamaIndex synthesises read queries; it computes no before/after diff of a write. nlqdb diff-previews writes and DDL before they apply. | ||
| Plan cache (repeat questions skip the LLM) | Caching identical questions is yours to build in LlamaIndex. nlqdb caches the compiled plan so a repeated question returns without another model call. | ||
| HTML embed element + anonymous try | nlqdb ships `<nlq-data>` and anonymous mode (a first answer before sign-in); a LlamaIndex query engine is Python you deploy and host yourself. | ||
| Broad RAG / retrieval framework (documents, vectors, agents) | LlamaIndex's reach across loaders, indexes, and retrievers is its moat; nlqdb is Postgres-first NL→SQL, not a general data framework. |
shipped · partial · not shipped
When to choose nlqdb
- You want NL→SQL working today without assembling a query engine, validator, and eval stack.
- You want the compiled SQL shown with every answer so you can audit the grain.
- You want a Postgres provisioned from English — no database to stand up first.
- You want destructive writes and migrations diff-previewed before they apply.
When to choose LlamaIndex
- You want full control of the retrieval, prompt, and query-engine internals.
- You're already in the LlamaIndex ecosystem building RAG over other data.
- You need it free, open source, and self-hosted against a database you run.
- You want to combine SQL with vector/document retrieval in one framework.
Questions buyers ask
- Is nlqdb a replacement for LlamaIndex?
- Only for the NL→SQL job. LlamaIndex's text-to-SQL is a query engine you assemble — schema retrieval, model, prompt, guardrails — over a database you already run. nlqdb is the hosted pipeline that does NL→SQL for you, provisions the Postgres, shows the SQL, and diff-previews writes. One is build-it-yourself; the other is embed-and-go.
- Can I use LlamaIndex and nlqdb together?
- Yes. A LlamaIndex agent or pipeline can call nlqdb as one tool — letting nlqdb own the database and the NL→SQL step while LlamaIndex handles document retrieval, vectors, and wider orchestration. You get nlqdb's validated, SQL-shown answers without hand-building the SQL query engine.
- Does LlamaIndex provision a database for me?
- No. `NLSQLTableQueryEngine` wraps a SQLAlchemy engine over a database you've already created and configured; it reads schema and runs SQL against it. nlqdb materialises a Postgres from your first English goal, so there's nothing to stand up before the first question.
- What do I have to build myself with LlamaIndex's text-to-SQL?
- The model choice, the prompt, schema retrieval (a TableIndex when the schema overflows context), SQL guardrails, caching, deployment, and evaluation. That flexibility is the point. nlqdb ships those as a managed pipeline, so you trade tunability for not maintaining the stack.
- Is running LlamaIndex's generated SQL safe by default?
- The LlamaIndex docs explicitly warn that executing arbitrary generated SQL is a security risk and recommend restricted roles, read-only databases, and sandboxing — the query tool runs whatever the model emits. nlqdb validates every query against a read-only allowlist and fails closed, and diff-previews any write before it applies.
Try nlqdb in 30 seconds
No sign-in. The anonymous database lasts 72 hours; adopt it with one click if you keep it.
Start with a goal →