Revert "Revert "Add custom JSON mode generation option for Gemini language mo…" (#84)

This reverts commit eca7b8f.

Author: mattt

Sources/AnyLanguageModel/Models/GeminiLanguageModel.swift

@@ -88,17 +88,47 @@ public struct GeminiLanguageModel: LanguageModel {
         /// (URL context), code execution, and location services (Google Maps).
         public var serverTools: [ServerTool]?
 
+        /// Configures JSON mode for structured output.
+        ///
+        /// Use this type to enable JSON mode,
+        /// which constrains the model to output a valid JSON.
+        /// Optionally provide a schema for typed JSON output.
+        public enum JSONMode: Sendable, Hashable, ExpressibleByBooleanLiteral {
+            /// JSON mode is disabled (default text output).
+            case disabled
+
+            /// JSON mode is enabled without a schema constraint.
+            case enabled
+
+            /// JSON mode is enabled with a schema constraint for typed output.
+            case schema(JSONSchema)
+
+            public init(booleanLiteral value: Bool) {
+                self = value ? .enabled : .disabled
+            }
+        }
+
+        /// The JSON mode configuration for structured output.
+        ///
+        /// When set to `.enabled`, the model will output valid JSON.
+        /// When set to `.schema(_:)`, the model will output JSON
+        /// conforming to the provided schema.
+        public var jsonMode: JSONMode?
+
         /// Creates custom generation options for Gemini models.
         ///
         /// - Parameters:
         ///   - thinking: The thinking mode configuration. When `nil`, uses the model's default.
         ///   - serverTools: Server-side tools to enable. When `nil`, uses the model's default.
+        ///   - jsonMode: The JSON mode configuration. When `nil`, uses the model's default.
         public init(
             thinking: Thinking? = nil,
-            serverTools: [ServerTool]? = nil
+            serverTools: [ServerTool]? = nil,
+            jsonMode: JSONMode? = nil
         ) {
             self.thinking = thinking
             self.serverTools = serverTools
+            self.jsonMode = jsonMode
         }
     }
 
@@ -240,6 +270,7 @@ public struct GeminiLanguageModel: LanguageModel {
         let customOptions = options[custom: GeminiLanguageModel.self]
         let effectiveThinking = customOptions?.thinking ?? _thinking
         let effectiveServerTools = customOptions?.serverTools ?? _serverTools
+        let effectiveJsonMode = customOptions?.jsonMode
 
         let url =
             baseURL
@@ -257,7 +288,8 @@ public struct GeminiLanguageModel: LanguageModel {
                 contents: transcript.toGeminiContent(),
                 tools: geminiTools,
                 options: options,
-                thinking: effectiveThinking
+                thinking: effectiveThinking,
+                jsonMode: effectiveJsonMode
             )
 
             let body = try JSONEncoder().encode(params)
@@ -326,6 +358,7 @@ public struct GeminiLanguageModel: LanguageModel {
         let customOptions = options[custom: GeminiLanguageModel.self]
         let effectiveThinking = customOptions?.thinking ?? _thinking
         let effectiveServerTools = customOptions?.serverTools ?? _serverTools
+        let effectiveJsonMode = customOptions?.jsonMode
 
         var streamURL =
             baseURL
@@ -346,7 +379,8 @@ public struct GeminiLanguageModel: LanguageModel {
                         contents: session.transcript.toGeminiContent(),
                         tools: geminiTools,
                         options: options,
-                        thinking: effectiveThinking
+                        thinking: effectiveThinking,
+                        jsonMode: effectiveJsonMode
                     )
 
                     let body = try JSONEncoder().encode(params)
@@ -431,7 +465,8 @@ private func createGenerateContentParams(
     contents: [GeminiContent],
     tools: [GeminiTool]?,
     options: GenerationOptions,
-    thinking: GeminiLanguageModel.CustomGenerationOptions.Thinking
+    thinking: GeminiLanguageModel.CustomGenerationOptions.Thinking,
+    jsonMode: GeminiLanguageModel.CustomGenerationOptions.JSONMode?
 ) throws -> [String: JSONValue] {
     var params: [String: JSONValue] = [
         "contents": try JSONValue(contents)
@@ -471,6 +506,18 @@ private func createGenerateContentParams(
     }
     generationConfig["thinkingConfig"] = .object(thinkingConfig)
 
+    if let jsonMode {
+        switch jsonMode {
+        case .disabled:
+            break
+        case .enabled:
+            generationConfig["responseMimeType"] = .string("application/json")
+        case .schema(let schema):
+            generationConfig["responseMimeType"] = .string("application/json")
+            generationConfig["responseSchema"] = try JSONValue(schema)
+        }
+    }
+
     if !generationConfig.isEmpty {
         params["generationConfig"] = .object(generationConfig)
     }

Tests/AnyLanguageModelTests/GeminiLanguageModelTests.swift

@@ -1,4 +1,5 @@
 import Foundation
+import JSONSchema
 import Testing
 
 @testable import AnyLanguageModel
@@ -134,4 +135,46 @@ struct GeminiLanguageModelTests {
         )
         #expect(!response.content.isEmpty)
     }
+
+    @Test func jsonModeEnabled() async throws {
+        let session = LanguageModelSession(model: model)
+
+        var options = GenerationOptions()
+        options[custom: GeminiLanguageModel.self] = .init(
+            thinking: .disabled,
+            jsonMode: true
+        )
+
+        let response = try await session.respond(
+            to: "Return a JSON object with a 'greeting' key and value 'hello'",
+            options: options
+        )
+        #expect(response.content.contains("greeting"))
+        #expect(response.content.contains("hello"))
+    }
+
+    @Test func jsonModeWithSchema() async throws {
+        let session = LanguageModelSession(model: model)
+
+        let schema = JSONSchema.object(
+            properties: [
+                "name": .string(),
+                "age": .integer(),
+            ],
+            required: ["name", "age"]
+        )
+
+        var options = GenerationOptions()
+        options[custom: GeminiLanguageModel.self] = .init(
+            thinking: .disabled,
+            jsonMode: .schema(schema)
+        )
+
+        let response = try await session.respond(
+            to: "Generate a person with name 'Alice' and age 30",
+            options: options
+        )
+        #expect(response.content.contains("Alice"))
+        #expect(response.content.contains("30"))
+    }
 }