feat: allow schema inference from collection validators MCP-603

[nirinchev]

Description

Adds convertMongoDBJSONSchemaToSimplified helper that reads the json schema validator and constructs simplified schema from it instead of using sampling.

[Copilot]

Copilot review overview

🟡 Changes recommended

Several conversions report types that the validator does not permit or omit valid types it accepts.

Review effort: Balanced
Findings: 1 · 5

Open (6)

• Correctly intersect sibling schemas with anyOf, oneOf, and allOf · New
• Map bsonType number to all BSON numeric types · New
• Do not infer Array type from additionalItems alone · New
• Preserve Document type unless DBRef shape is guaranteed · New
• Ignore additionalItems for single-schema items · New
• Broaden MongoDBJSONSchema to support standard JSON Schema keywords · New

What changed in this PR

Adds conversion of MongoDB collection validators into simplified schemas without document sampling.

Changes:

• Adds BSON/JSON Schema type conversion, unions, arrays, documents, encryption, and DBRef handling.
• Exports and documents the converter.
• Adds comprehensive tests and updates schema types.

File	Description
packages/​mongodb-schema/​src/​types.ts	Extends validator schema types.
packages/​mongodb-schema/​src/​schema-converters/​mongodb-to-simplified.ts	Implements validator conversion.
packages/​mongodb-schema/​src/​schema-converters/​mongodb-to-simplified.test.ts	Tests conversion behavior.
packages/​mongodb-schema/​src/​schema-analyzer.ts	Aligns simplified BSON type names.
packages/​mongodb-schema/​src/​index.ts	Exports the converter.
packages/​mongodb-schema/​README.md	Documents validator-based inference.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟡 Changes recommended

Array unions, tuple additional items, DBRef detection, and the exported BSON type union currently produce inaccurate results.

Review effort: Balanced
Findings: 3

Open (3)

• Preserve unconstrained array members when merging union branches · New
• Treat absent or true additionalItems as unconstrained tuple members · New
• Broaden MongoDBJSONSchema to support standard JSON Schema keywords

Resolved since last review (5)

• Correctly intersect sibling schemas with anyOf, oneOf, and allOf
• Ignore additionalItems for single-schema items
• Preserve Document type unless DBRef shape is guaranteed
• Do not infer Array type from additionalItems alone
• Map bsonType number to all BSON numeric types

Previously missed (2)

In code that hasn't changed since last review

Include Number, RegExp, and Symbol in the public schema type union

packages/​mongodb-schema/​src/​schema-analyzer.ts:131

This newly exported public type is knowingly narrower than the values returned by getSimplifiedSchema: ordinary JavaScript numbers, regular expressions, and symbols produce Number, RegExp, and Symbol. Consumers can therefore write an apparently exhaustive mapping that compiles but fails on valid simplified schemas (the test generator already has to patch in Number). Include all runtime values in the public union and update downstream mappings accordingly.

Detect DBRefs from raw properties without type information

packages/​mongodb-schema/​src/​schema-converters/​mongodb-to-simplified.ts:335

DBRef detection is based on fields, but collectFields deliberately omits properties whose subschema has no type information. A valid DBRef-shaped validator commonly uses $id: {} (the repository's standard DBRef definition does this), so even with both names required this path returns Document instead of the DBRef sampling would observe. Detect the two property names from the raw properties map, while continuing to use fields for the simplified document shape.