graphile-llm 0.2.0

package/__tests__/graphile-llm.test.js

@@ -0,0 +1,721 @@
+"use strict";
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+const path_1 = require("path");
+const ollama_1 = __importDefault(require("@agentic-kit/ollama"));
+const graphile_test_1 = require("graphile-test");
+const graphile_connection_filter_1 = require("graphile-connection-filter");
+const vector_codec_1 = require("graphile-search/codecs/vector-codec");
+const plugin_1 = require("graphile-search/plugin");
+const pgvector_1 = require("graphile-search/adapters/pgvector");
+const llm_module_plugin_1 = require("../../src/plugins/llm-module-plugin");
+const text_search_plugin_1 = require("../../src/plugins/text-search-plugin");
+const text_mutation_plugin_1 = require("../../src/plugins/text-mutation-plugin");
+const rag_plugin_1 = require("../../src/plugins/rag-plugin");
+const embedder_1 = require("../../src/embedder");
+const chat_1 = require("../../src/chat");
+// ─── @agentic-kit/ollama client ─────────────────────────────────────────────
+const ollamaClient = new ollama_1.default('http://localhost:11434');
+async function ensureNomicModel() {
+    const models = await ollamaClient.listModels();
+    const hasModel = models.some((m) => m.includes('nomic-embed-text'));
+    if (!hasModel) {
+        console.log('Pulling nomic-embed-text model...');
+        await ollamaClient.pullModel('nomic-embed-text');
+    }
+}
+// =============================================================================
+// Suite 1: Embedder abstraction unit tests
+// =============================================================================
+describe('Embedder abstraction', () => {
+    describe('buildEmbedder()', () => {
+        it('returns an EmbedderFunction for ollama provider', () => {
+            const embedder = (0, embedder_1.buildEmbedder)({
+                provider: 'ollama',
+                model: 'nomic-embed-text',
+                baseUrl: 'http://localhost:11434',
+            });
+            expect(embedder).not.toBeNull();
+            expect(typeof embedder).toBe('function');
+        });
+        it('returns null for unknown provider', () => {
+            const embedder = (0, embedder_1.buildEmbedder)({
+                provider: 'unknown-provider',
+            });
+            expect(embedder).toBeNull();
+        });
+        it('uses default model and baseUrl for ollama when not specified', () => {
+            const embedder = (0, embedder_1.buildEmbedder)({ provider: 'ollama' });
+            expect(embedder).not.toBeNull();
+            expect(typeof embedder).toBe('function');
+        });
+    });
+    describe('buildEmbedderFromModule()', () => {
+        it('builds embedder from LlmModuleData', () => {
+            const moduleData = {
+                embedding_provider: 'ollama',
+                embedding_model: 'nomic-embed-text',
+                embedding_base_url: 'http://localhost:11434',
+            };
+            const embedder = (0, embedder_1.buildEmbedderFromModule)(moduleData);
+            expect(embedder).not.toBeNull();
+            expect(typeof embedder).toBe('function');
+        });
+        it('returns null for unsupported provider in module data', () => {
+            const moduleData = {
+                embedding_provider: 'unsupported',
+            };
+            const embedder = (0, embedder_1.buildEmbedderFromModule)(moduleData);
+            expect(embedder).toBeNull();
+        });
+    });
+    describe('buildEmbedderFromEnv()', () => {
+        const originalEnv = process.env;
+        afterEach(() => {
+            process.env = originalEnv;
+        });
+        it('returns null when EMBEDDER_PROVIDER is not set', () => {
+            process.env = { ...originalEnv };
+            delete process.env.EMBEDDER_PROVIDER;
+            const embedder = (0, embedder_1.buildEmbedderFromEnv)();
+            expect(embedder).toBeNull();
+        });
+        it('builds embedder from environment variables', () => {
+            process.env = {
+                ...originalEnv,
+                EMBEDDER_PROVIDER: 'ollama',
+                EMBEDDER_MODEL: 'nomic-embed-text',
+                EMBEDDER_BASE_URL: 'http://localhost:11434',
+            };
+            const embedder = (0, embedder_1.buildEmbedderFromEnv)();
+            expect(embedder).not.toBeNull();
+            expect(typeof embedder).toBe('function');
+        });
+    });
+});
+// =============================================================================
+// Suite 2: Schema enrichment — plugin adds text fields to GraphQL schema
+// Requires PostgreSQL + pgvector. Tests WILL fail if database is unavailable.
+// =============================================================================
+describe('graphile-llm schema enrichment', () => {
+    let db;
+    let teardown;
+    let query;
+    beforeAll(async () => {
+        const unifiedPlugin = (0, plugin_1.createUnifiedSearchPlugin)({
+            adapters: [(0, pgvector_1.createPgvectorAdapter)()],
+        });
+        const testPreset = {
+            extends: [(0, graphile_connection_filter_1.ConnectionFilterPreset)()],
+            plugins: [
+                // Search infrastructure (provides VectorNearbyInput)
+                vector_codec_1.VectorCodecPlugin,
+                unifiedPlugin,
+                // LLM plugins under test
+                (0, llm_module_plugin_1.createLlmModulePlugin)({
+                    defaultEmbedder: {
+                        provider: 'ollama',
+                        model: 'nomic-embed-text',
+                        baseUrl: 'http://localhost:11434',
+                    },
+                }),
+                (0, text_search_plugin_1.createLlmTextSearchPlugin)(),
+                (0, text_mutation_plugin_1.createLlmTextMutationPlugin)(),
+            ],
+        };
+        const connections = await (0, graphile_test_1.getConnections)({
+            schemas: ['llm_test'],
+            preset: testPreset,
+            useRoot: true,
+            authRole: 'postgres',
+        }, [graphile_test_1.seed.sqlfile([(0, path_1.join)(__dirname, './setup.sql')])]);
+        db = connections.db;
+        teardown = connections.teardown;
+        query = connections.query;
+    });
+    afterAll(async () => {
+        if (teardown) {
+            await teardown();
+        }
+    });
+    beforeEach(async () => {
+        await db.beforeEach();
+    });
+    afterEach(async () => {
+        await db.afterEach();
+    });
+    // ─── VectorNearbyInput text field ────────────────────────────────────────
+    describe('VectorNearbyInput text field', () => {
+        it('adds text field to VectorNearbyInput type', async () => {
+            const result = await query(`
+        query {
+          __type(name: "VectorNearbyInput") {
+            inputFields {
+              name
+              type { name }
+            }
+          }
+        }
+      `);
+            expect(result.errors).toBeUndefined();
+            const inputType = result.data?.__type;
+            expect(inputType).toBeDefined();
+            const fieldNames = inputType.inputFields.map((f) => f.name);
+            // Original fields from graphile-search
+            expect(fieldNames).toContain('vector');
+            expect(fieldNames).toContain('metric');
+            expect(fieldNames).toContain('distance');
+            // New field from LlmTextSearchPlugin
+            expect(fieldNames).toContain('text');
+            // text field should be a String
+            const textField = inputType.inputFields.find((f) => f.name === 'text');
+            expect(textField.type.name).toBe('String');
+        });
+        it('still allows vector-based queries (existing behavior unchanged)', async () => {
+            const result = await query(`
+        query {
+          allArticles(where: {
+            vectorEmbedding: {
+              vector: [1, 0, 0]
+              metric: COSINE
+            }
+          }) {
+            nodes {
+              title
+              embeddingVectorDistance
+            }
+          }
+        }
+      `);
+            expect(result.errors).toBeUndefined();
+            const nodes = result.data?.allArticles?.nodes;
+            expect(nodes).toBeDefined();
+            expect(nodes.length).toBeGreaterThan(0);
+            for (const node of nodes) {
+                expect(typeof node.embeddingVectorDistance).toBe('number');
+                expect(node.embeddingVectorDistance).toBeGreaterThanOrEqual(0);
+            }
+        });
+    });
+    // ─── Mutation text companion fields ──────────────────────────────────────
+    describe('Mutation text companion fields', () => {
+        it('adds embeddingText field to CreateArticleInput', async () => {
+            const result = await query(`
+        query {
+          __type(name: "CreateArticleInput") {
+            inputFields {
+              name
+              type { name }
+            }
+          }
+        }
+      `);
+            expect(result.errors).toBeUndefined();
+            const inputType = result.data?.__type;
+            expect(inputType).toBeDefined();
+            const fieldNames = inputType.inputFields.map((f) => f.name);
+            // Original embedding field
+            expect(fieldNames).toContain('embedding');
+            // Companion text field from LlmTextMutationPlugin
+            expect(fieldNames).toContain('embeddingText');
+            const textField = inputType.inputFields.find((f) => f.name === 'embeddingText');
+            expect(textField).toBeDefined();
+            expect(textField.type.name).toBe('String');
+        });
+        it('adds embeddingText field to UpdateArticleInput (patch)', async () => {
+            const result = await query(`
+        query {
+          __type(name: "ArticlePatch") {
+            inputFields {
+              name
+              type { name }
+            }
+          }
+        }
+      `);
+            expect(result.errors).toBeUndefined();
+            const inputType = result.data?.__type;
+            expect(inputType).toBeDefined();
+            const fieldNames = inputType.inputFields.map((f) => f.name);
+            expect(fieldNames).toContain('embeddingText');
+            const textField = inputType.inputFields.find((f) => f.name === 'embeddingText');
+            expect(textField).toBeDefined();
+            expect(textField.type.name).toBe('String');
+        });
+    });
+});
+// =============================================================================
+// Suite 3: Real Ollama embedding via @agentic-kit/ollama
+// Requires Ollama running with nomic-embed-text. Tests WILL fail if unavailable.
+// =============================================================================
+describe('graphile-llm with real Ollama embedding', () => {
+    beforeAll(async () => {
+        await ensureNomicModel();
+    });
+    it('should embed text to a real vector via Ollama nomic-embed-text', async () => {
+        const embedder = (0, embedder_1.buildEmbedder)({
+            provider: 'ollama',
+            model: 'nomic-embed-text',
+            baseUrl: 'http://localhost:11434',
+        });
+        expect(embedder).not.toBeNull();
+        const vector = await embedder('Machine learning is transforming AI');
+        // nomic-embed-text produces 768-dimensional vectors
+        expect(Array.isArray(vector)).toBe(true);
+        expect(vector.length).toBe(768);
+        // All elements should be numbers
+        for (const v of vector) {
+            expect(typeof v).toBe('number');
+            expect(Number.isFinite(v)).toBe(true);
+        }
+        // Vector should not be all zeros
+        const magnitude = Math.sqrt(vector.reduce((sum, v) => sum + v * v, 0));
+        expect(magnitude).toBeGreaterThan(0);
+    });
+    it('should produce different vectors for semantically different text', async () => {
+        const embedder = (0, embedder_1.buildEmbedder)({
+            provider: 'ollama',
+            model: 'nomic-embed-text',
+            baseUrl: 'http://localhost:11434',
+        });
+        expect(embedder).not.toBeNull();
+        const [vecA, vecB] = await Promise.all([
+            embedder('Artificial intelligence and machine learning'),
+            embedder('Cooking recipes for Italian pasta dishes'),
+        ]);
+        expect(vecA.length).toBe(768);
+        expect(vecB.length).toBe(768);
+        // Compute cosine similarity
+        let dotProduct = 0;
+        let normA = 0;
+        let normB = 0;
+        for (let i = 0; i < vecA.length; i++) {
+            dotProduct += vecA[i] * vecB[i];
+            normA += vecA[i] * vecA[i];
+            normB += vecB[i] * vecB[i];
+        }
+        const cosineSimilarity = dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));
+        // Semantically different texts should have lower similarity
+        expect(cosineSimilarity).toBeLessThan(0.95);
+    });
+    it('should produce similar vectors for semantically similar text', async () => {
+        const embedder = (0, embedder_1.buildEmbedder)({
+            provider: 'ollama',
+            model: 'nomic-embed-text',
+            baseUrl: 'http://localhost:11434',
+        });
+        expect(embedder).not.toBeNull();
+        const [vecA, vecB] = await Promise.all([
+            embedder('Machine learning and artificial intelligence'),
+            embedder('AI and ML are subfields of computer science'),
+        ]);
+        expect(vecA.length).toBe(768);
+        expect(vecB.length).toBe(768);
+        // Compute cosine similarity
+        let dotProduct = 0;
+        let normA = 0;
+        let normB = 0;
+        for (let i = 0; i < vecA.length; i++) {
+            dotProduct += vecA[i] * vecB[i];
+            normA += vecA[i] * vecA[i];
+            normB += vecB[i] * vecB[i];
+        }
+        const cosineSimilarity = dotProduct / (Math.sqrt(normA) * Math.sqrt(normB));
+        // Semantically similar texts should have high similarity
+        expect(cosineSimilarity).toBeGreaterThan(0.5);
+    });
+    it('should produce embeddings via @agentic-kit/ollama OllamaClient directly', async () => {
+        const vector = await ollamaClient.generateEmbedding('Testing the agentic-kit Ollama client directly', 'nomic-embed-text');
+        expect(Array.isArray(vector)).toBe(true);
+        expect(vector.length).toBe(768);
+        for (const v of vector) {
+            expect(typeof v).toBe('number');
+            expect(Number.isFinite(v)).toBe(true);
+        }
+    });
+});
+// =============================================================================
+// Suite 4: Chat completion abstraction unit tests
+// =============================================================================
+describe('Chat completion abstraction', () => {
+    describe('buildChatCompleter()', () => {
+        it('returns a ChatFunction for ollama provider', () => {
+            const chat = (0, chat_1.buildChatCompleter)({
+                provider: 'ollama',
+                model: 'llama3',
+                baseUrl: 'http://localhost:11434',
+            });
+            expect(chat).not.toBeNull();
+            expect(typeof chat).toBe('function');
+        });
+        it('returns null for unknown provider', () => {
+            const chat = (0, chat_1.buildChatCompleter)({
+                provider: 'unknown-provider',
+            });
+            expect(chat).toBeNull();
+        });
+        it('uses default model and baseUrl for ollama when not specified', () => {
+            const chat = (0, chat_1.buildChatCompleter)({ provider: 'ollama' });
+            expect(chat).not.toBeNull();
+            expect(typeof chat).toBe('function');
+        });
+    });
+    describe('buildChatCompleterFromModule()', () => {
+        it('builds chat completer from LlmModuleData with chat_provider', () => {
+            const moduleData = {
+                embedding_provider: 'ollama',
+                chat_provider: 'ollama',
+                chat_model: 'llama3',
+                chat_base_url: 'http://localhost:11434',
+            };
+            const chat = (0, chat_1.buildChatCompleterFromModule)(moduleData);
+            expect(chat).not.toBeNull();
+            expect(typeof chat).toBe('function');
+        });
+        it('returns null when chat_provider is not set', () => {
+            const moduleData = {
+                embedding_provider: 'ollama',
+            };
+            const chat = (0, chat_1.buildChatCompleterFromModule)(moduleData);
+            expect(chat).toBeNull();
+        });
+    });
+    describe('buildChatCompleterFromEnv()', () => {
+        const originalEnv = process.env;
+        afterEach(() => {
+            process.env = originalEnv;
+        });
+        it('returns null when CHAT_PROVIDER is not set', () => {
+            process.env = { ...originalEnv };
+            delete process.env.CHAT_PROVIDER;
+            const chat = (0, chat_1.buildChatCompleterFromEnv)();
+            expect(chat).toBeNull();
+        });
+        it('builds chat completer from environment variables', () => {
+            process.env = {
+                ...originalEnv,
+                CHAT_PROVIDER: 'ollama',
+                CHAT_MODEL: 'llama3',
+                CHAT_BASE_URL: 'http://localhost:11434',
+            };
+            const chat = (0, chat_1.buildChatCompleterFromEnv)();
+            expect(chat).not.toBeNull();
+            expect(typeof chat).toBe('function');
+        });
+    });
+});
+// =============================================================================
+// Suite 5: RAG plugin schema enrichment
+// Requires PostgreSQL + pgvector. Tests WILL fail if database is unavailable.
+// =============================================================================
+/**
+ * Smart tag injection plugin for test tables.
+ * Injects @hasChunks on the articles codec so the RAG plugin can discover it.
+ */
+function makeTestSmartTagsPlugin(tagsByTable) {
+    return {
+        name: 'TestSmartTagsPlugin',
+        version: '1.0.0',
+        schema: {
+            hooks: {
+                init: {
+                    before: ['UnifiedSearchPlugin', 'LlmRagPlugin'],
+                    callback(_, build) {
+                        for (const codec of Object.values(build.input.pgRegistry.pgCodecs)) {
+                            const c = codec;
+                            if (!c.attributes || !c.name)
+                                continue;
+                            const tags = tagsByTable[c.name];
+                            if (!tags)
+                                continue;
+                            if (!c.extensions)
+                                c.extensions = {};
+                            if (!c.extensions.tags)
+                                c.extensions.tags = {};
+                            Object.assign(c.extensions.tags, tags);
+                        }
+                        return _;
+                    },
+                },
+            },
+        },
+    };
+}
+describe('RAG plugin schema enrichment', () => {
+    let db;
+    let teardown;
+    let query;
+    beforeAll(async () => {
+        const unifiedPlugin = (0, plugin_1.createUnifiedSearchPlugin)({
+            adapters: [(0, pgvector_1.createPgvectorAdapter)()],
+        });
+        const smartTagsPlugin = makeTestSmartTagsPlugin({
+            articles: {
+                hasChunks: {
+                    chunksTable: 'articles_chunks',
+                    parentFk: 'parent_id',
+                    parentPk: 'id',
+                    embeddingField: 'embedding',
+                    contentField: 'content',
+                },
+            },
+        });
+        // Mock embedder that returns a fixed 3-dim vector
+        const mockEmbedder = async (_text) => [1, 0, 0];
+        // Mock chat completer that returns a canned response
+        const mockChatCompleter = async (messages) => {
+            const userMessage = messages.find((m) => m.role === 'user');
+            return `Mock answer for: ${userMessage?.content || 'unknown'}`;
+        };
+        const testPreset = {
+            extends: [(0, graphile_connection_filter_1.ConnectionFilterPreset)()],
+            plugins: [
+                vector_codec_1.VectorCodecPlugin,
+                unifiedPlugin,
+                smartTagsPlugin,
+                (0, llm_module_plugin_1.createLlmModulePlugin)({
+                    defaultEmbedder: {
+                        provider: 'ollama',
+                        model: 'nomic-embed-text',
+                        baseUrl: 'http://localhost:11434',
+                    },
+                }),
+                (0, text_search_plugin_1.createLlmTextSearchPlugin)(),
+                (0, text_mutation_plugin_1.createLlmTextMutationPlugin)(),
+                (0, rag_plugin_1.createLlmRagPlugin)(),
+            ],
+        };
+        // Override the embedder and chat completer on the build context
+        // by wrapping the LlmModulePlugin's build hook
+        const overridePlugin = {
+            name: 'TestOverridePlugin',
+            version: '1.0.0',
+            after: ['LlmModulePlugin'],
+            schema: {
+                hooks: {
+                    build(build) {
+                        return build.extend(build, {
+                            llmEmbedder: mockEmbedder,
+                            llmChatCompleter: mockChatCompleter,
+                        }, 'TestOverridePlugin overriding embedder and chat completer');
+                    },
+                },
+            },
+        };
+        const connections = await (0, graphile_test_1.getConnections)({
+            schemas: ['llm_test'],
+            preset: {
+                ...testPreset,
+                plugins: [...testPreset.plugins, overridePlugin],
+            },
+            useRoot: true,
+            authRole: 'postgres',
+        }, [graphile_test_1.seed.sqlfile([(0, path_1.join)(__dirname, './setup.sql')])]);
+        db = connections.db;
+        teardown = connections.teardown;
+        query = connections.query;
+    });
+    afterAll(async () => {
+        if (teardown) {
+            await teardown();
+        }
+    });
+    beforeEach(async () => {
+        await db.beforeEach();
+    });
+    afterEach(async () => {
+        await db.afterEach();
+    });
+    it('adds ragQuery field to the Query type', async () => {
+        const result = await query(`
+      query {
+        __type(name: "Query") {
+          fields {
+            name
+          }
+        }
+      }
+    `);
+        expect(result.errors).toBeUndefined();
+        const queryType = result.data?.__type;
+        expect(queryType).toBeDefined();
+        const fieldNames = queryType.fields.map((f) => f.name);
+        expect(fieldNames).toContain('ragQuery');
+    });
+    it('adds embedText field to the Query type', async () => {
+        const result = await query(`
+      query {
+        __type(name: "Query") {
+          fields {
+            name
+          }
+        }
+      }
+    `);
+        expect(result.errors).toBeUndefined();
+        const fieldNames = result.data.__type.fields.map((f) => f.name);
+        expect(fieldNames).toContain('embedText');
+    });
+    it('ragQuery returns RagResponse type with answer and sources', async () => {
+        const result = await query(`
+      query {
+        __type(name: "RagResponse") {
+          fields {
+            name
+            type { name kind }
+          }
+        }
+      }
+    `);
+        expect(result.errors).toBeUndefined();
+        const ragResponseType = result.data?.__type;
+        expect(ragResponseType).toBeDefined();
+        const fieldNames = ragResponseType.fields.map((f) => f.name);
+        expect(fieldNames).toContain('answer');
+        expect(fieldNames).toContain('sources');
+        expect(fieldNames).toContain('tokensUsed');
+    });
+    it('ragQuery executes and returns mock answer with sources', async () => {
+        const result = await query(`
+      query {
+        ragQuery(prompt: "What is machine learning?", contextLimit: 3) {
+          answer
+          sources {
+            content
+            similarity
+            tableName
+            parentId
+          }
+          tokensUsed
+        }
+      }
+    `);
+        expect(result.errors).toBeUndefined();
+        const rag = result.data?.ragQuery;
+        expect(rag).toBeDefined();
+        // Mock chat completer returns a canned response
+        expect(rag.answer).toContain('Mock answer for');
+        expect(rag.answer).toContain('What is machine learning?');
+        // Sources should come from the chunks table
+        expect(rag.sources.length).toBeGreaterThan(0);
+        expect(rag.sources.length).toBeLessThanOrEqual(3);
+        for (const source of rag.sources) {
+            expect(typeof source.content).toBe('string');
+            expect(source.content.length).toBeGreaterThan(0);
+            expect(typeof source.similarity).toBe('number');
+            expect(source.similarity).toBeGreaterThanOrEqual(0);
+            expect(source.similarity).toBeLessThanOrEqual(1);
+            expect(source.tableName).toBe('articles');
+            expect(source.parentId).toBeTruthy();
+        }
+    });
+    it('embedText executes and returns vector with dimensions', async () => {
+        const result = await query(`
+      query {
+        embedText(text: "test embedding") {
+          vector
+          dimensions
+        }
+      }
+    `);
+        expect(result.errors).toBeUndefined();
+        const embed = result.data?.embedText;
+        expect(embed).toBeDefined();
+        // Mock embedder returns [1, 0, 0]
+        expect(embed.vector).toEqual([1, 0, 0]);
+        expect(embed.dimensions).toBe(3);
+    });
+});
+// =============================================================================
+// Suite 6: GraphileLlmPreset toggle tests
+// =============================================================================
+describe('GraphileLlmPreset toggles', () => {
+    it('enableRag=false excludes RAG plugin (no ragQuery field)', async () => {
+        const { GraphileLlmPreset } = await Promise.resolve().then(() => __importStar(require('../../src/preset')));
+        const preset = GraphileLlmPreset({
+            enableRag: false,
+        });
+        const pluginNames = preset.plugins.map((p) => p.name);
+        expect(pluginNames).not.toContain('LlmRagPlugin');
+    });
+    it('enableRag=true includes RAG plugin', async () => {
+        const { GraphileLlmPreset } = await Promise.resolve().then(() => __importStar(require('../../src/preset')));
+        const preset = GraphileLlmPreset({
+            enableRag: true,
+        });
+        const pluginNames = preset.plugins.map((p) => p.name);
+        expect(pluginNames).toContain('LlmRagPlugin');
+    });
+    it('enableTextSearch=false excludes text search plugin', async () => {
+        const { GraphileLlmPreset } = await Promise.resolve().then(() => __importStar(require('../../src/preset')));
+        const preset = GraphileLlmPreset({
+            enableTextSearch: false,
+        });
+        const pluginNames = preset.plugins.map((p) => p.name);
+        expect(pluginNames).not.toContain('LlmTextSearchPlugin');
+        // Module plugin is always included
+        expect(pluginNames).toContain('LlmModulePlugin');
+    });
+    it('enableTextMutations=false excludes text mutation plugin', async () => {
+        const { GraphileLlmPreset } = await Promise.resolve().then(() => __importStar(require('../../src/preset')));
+        const preset = GraphileLlmPreset({
+            enableTextMutations: false,
+        });
+        const pluginNames = preset.plugins.map((p) => p.name);
+        expect(pluginNames).not.toContain('LlmTextMutationPlugin');
+    });
+    it('all toggles false leaves only LlmModulePlugin', async () => {
+        const { GraphileLlmPreset } = await Promise.resolve().then(() => __importStar(require('../../src/preset')));
+        const preset = GraphileLlmPreset({
+            enableTextSearch: false,
+            enableTextMutations: false,
+            enableRag: false,
+        });
+        const pluginNames = preset.plugins.map((p) => p.name);
+        expect(pluginNames).toEqual(['LlmModulePlugin']);
+    });
+    it('default options include text search and mutations but not RAG', async () => {
+        const { GraphileLlmPreset } = await Promise.resolve().then(() => __importStar(require('../../src/preset')));
+        const preset = GraphileLlmPreset();
+        const pluginNames = preset.plugins.map((p) => p.name);
+        expect(pluginNames).toContain('LlmModulePlugin');
+        expect(pluginNames).toContain('LlmTextSearchPlugin');
+        expect(pluginNames).toContain('LlmTextMutationPlugin');
+        expect(pluginNames).not.toContain('LlmRagPlugin');
+    });
+});