@polymorphism-tech/morph-spec 2.4.0 → 3.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (218) hide show
  1. package/CLAUDE.md +158 -26
  2. package/LICENSE +72 -72
  3. package/bin/detect-agents.js +225 -225
  4. package/bin/morph-spec.js +8 -0
  5. package/bin/render-template.js +302 -302
  6. package/bin/semantic-detect-agents.js +246 -246
  7. package/bin/validate-agents-skills.js +251 -251
  8. package/bin/validate-agents.js +69 -69
  9. package/bin/validate-phase.js +263 -263
  10. package/content/.azure/README.md +293 -293
  11. package/content/.azure/docs/azure-devops-setup.md +454 -454
  12. package/content/.azure/docs/branch-strategy.md +398 -398
  13. package/content/.azure/docs/local-development.md +515 -515
  14. package/content/.azure/pipelines/pipeline-variables.yml +34 -34
  15. package/content/.azure/pipelines/prod-pipeline.yml +319 -319
  16. package/content/.azure/pipelines/staging-pipeline.yml +234 -234
  17. package/content/.azure/pipelines/templates/build-dotnet.yml +75 -75
  18. package/content/.azure/pipelines/templates/deploy-app-service.yml +94 -94
  19. package/content/.azure/pipelines/templates/deploy-container-app.yml +120 -120
  20. package/content/.azure/pipelines/templates/infra-deploy.yml +90 -90
  21. package/content/.claude/commands/morph-archive.md +79 -79
  22. package/content/.claude/commands/morph-deploy.md +529 -0
  23. package/content/.claude/commands/morph-infra.md +209 -209
  24. package/content/.claude/commands/morph-preflight.md +227 -227
  25. package/content/.claude/commands/morph-troubleshoot.md +122 -122
  26. package/content/.claude/settings.local.json +15 -15
  27. package/content/.claude/skills/infra/azure-deploy-specialist.md +699 -0
  28. package/content/.claude/skills/level-0-meta/README.md +7 -0
  29. package/content/.claude/skills/{checklists → level-0-meta}/morph-checklist.md +117 -117
  30. package/content/.claude/skills/level-1-workflows/README.md +7 -0
  31. package/content/.claude/skills/{workflows → level-1-workflows}/morph-replicate.md +213 -213
  32. package/content/.claude/skills/{workflows → level-1-workflows}/phase-clarify.md +131 -131
  33. package/content/.claude/skills/{workflows → level-1-workflows}/phase-design.md +213 -205
  34. package/content/.claude/skills/{workflows → level-1-workflows}/phase-setup.md +106 -92
  35. package/content/.claude/skills/{workflows → level-1-workflows}/phase-tasks.md +164 -164
  36. package/content/.claude/skills/{workflows → level-1-workflows}/phase-uiux.md +169 -138
  37. package/content/.claude/skills/level-2-domains/README.md +14 -0
  38. package/content/.claude/skills/{specialists → level-2-domains/quality}/testing-specialist.md +126 -126
  39. package/content/.claude/skills/level-3-technologies/README.md +7 -0
  40. package/content/.claude/skills/level-4-patterns/README.md +7 -0
  41. package/content/.claude/skills/specialists/prompt-engineer.md +189 -0
  42. package/content/.claude/skills/specialists/seo-growth-hacker.md +320 -0
  43. package/content/.morph/.morphversion +5 -5
  44. package/content/.morph/archive/.gitkeep +25 -25
  45. package/content/.morph/config/agents.json +742 -358
  46. package/content/.morph/config/config.template.json +33 -0
  47. package/content/.morph/docs/STORY-DRIVEN-DEVELOPMENT.md +392 -392
  48. package/content/.morph/docs/workflows/enforcement-pipeline.md +668 -0
  49. package/content/.morph/examples/api-nextjs/README.md +241 -241
  50. package/content/.morph/examples/api-nextjs/contracts.ts +307 -307
  51. package/content/.morph/examples/api-nextjs/spec.md +399 -399
  52. package/content/.morph/examples/api-nextjs/tasks.md +168 -168
  53. package/content/.morph/examples/micro-saas/README.md +125 -125
  54. package/content/.morph/examples/micro-saas/contracts.cs +358 -358
  55. package/content/.morph/examples/micro-saas/decisions.md +246 -246
  56. package/content/.morph/examples/micro-saas/spec.md +236 -236
  57. package/content/.morph/examples/micro-saas/tasks.md +150 -150
  58. package/content/.morph/examples/multi-agent/README.md +309 -309
  59. package/content/.morph/examples/multi-agent/contracts.cs +433 -433
  60. package/content/.morph/examples/multi-agent/spec.md +479 -479
  61. package/content/.morph/examples/multi-agent/tasks.md +185 -185
  62. package/content/.morph/examples/scheduled-reports/decisions.md +158 -158
  63. package/content/.morph/examples/scheduled-reports/proposal.md +95 -95
  64. package/content/.morph/examples/scheduled-reports/spec.md +267 -267
  65. package/content/.morph/examples/state-v3.json +188 -188
  66. package/content/.morph/features/.gitkeep +25 -25
  67. package/content/.morph/hooks/README.md +158 -0
  68. package/content/.morph/hooks/pre-commit-all.sh +48 -48
  69. package/content/.morph/hooks/pre-commit-specs.sh +49 -49
  70. package/content/.morph/hooks/pre-commit-tests.sh +60 -60
  71. package/content/.morph/hooks/task-completed.js +73 -0
  72. package/content/.morph/hooks/teammate-idle.js +68 -0
  73. package/content/.morph/project.md +160 -160
  74. package/content/.morph/schemas/agent.schema.json +296 -296
  75. package/content/.morph/schemas/tasks.schema.json +220 -220
  76. package/content/.morph/specs/.gitkeep +20 -20
  77. package/content/.morph/standards/agent-teams-workflow.md +474 -0
  78. package/content/.morph/standards/coding.md +377 -377
  79. package/content/.morph/standards/fluent-ui-setup.md +590 -590
  80. package/content/.morph/standards/migration-guide.md +514 -514
  81. package/content/.morph/standards/passkeys-auth.md +423 -423
  82. package/content/.morph/standards/vector-search-rag.md +536 -536
  83. package/content/.morph/state.json +17 -17
  84. package/content/.morph/templates/CONTEXT-FEATURE.md +276 -0
  85. package/content/.morph/templates/CONTEXT.md +170 -0
  86. package/content/.morph/templates/FluentDesignTheme.cs +149 -149
  87. package/content/.morph/templates/MudTheme.cs +281 -281
  88. package/content/.morph/templates/clarify-questions.md +159 -159
  89. package/content/.morph/templates/component.razor +239 -239
  90. package/content/.morph/templates/contracts/Commands.cs +74 -74
  91. package/content/.morph/templates/contracts/Entities.cs +25 -25
  92. package/content/.morph/templates/contracts/Queries.cs +74 -74
  93. package/content/.morph/templates/contracts/README.md +74 -74
  94. package/content/.morph/templates/contracts.cs +217 -217
  95. package/content/.morph/templates/design-system.css +226 -226
  96. package/content/.morph/templates/infra/.dockerignore.example +89 -89
  97. package/content/.morph/templates/infra/Dockerfile.example +82 -82
  98. package/content/.morph/templates/infra/README.md +286 -286
  99. package/content/.morph/templates/infra/app-insights.bicep +63 -63
  100. package/content/.morph/templates/infra/app-service.bicep +164 -164
  101. package/content/.morph/templates/infra/azure-pipelines-deploy.yml +480 -0
  102. package/content/.morph/templates/infra/container-app-env.bicep +49 -49
  103. package/content/.morph/templates/infra/container-app.bicep +156 -156
  104. package/content/.morph/templates/infra/deploy-checklist.md +426 -426
  105. package/content/.morph/templates/infra/deploy.ps1 +229 -229
  106. package/content/.morph/templates/infra/deploy.sh +208 -208
  107. package/content/.morph/templates/infra/key-vault.bicep +91 -91
  108. package/content/.morph/templates/infra/main.bicep +189 -189
  109. package/content/.morph/templates/infra/parameters.dev.json +29 -29
  110. package/content/.morph/templates/infra/parameters.prod.json +29 -29
  111. package/content/.morph/templates/infra/parameters.staging.json +29 -29
  112. package/content/.morph/templates/infra/sql-database.bicep +103 -103
  113. package/content/.morph/templates/infra/storage.bicep +106 -106
  114. package/content/.morph/templates/integrations/asaas-client.cs +387 -387
  115. package/content/.morph/templates/integrations/asaas-webhook.cs +351 -351
  116. package/content/.morph/templates/integrations/azure-identity-config.cs +288 -288
  117. package/content/.morph/templates/integrations/clerk-config.cs +258 -258
  118. package/content/.morph/templates/job.cs +171 -171
  119. package/content/.morph/templates/migration.cs +83 -83
  120. package/content/.morph/templates/repository.cs +141 -141
  121. package/content/.morph/templates/saas/subscription.cs +347 -347
  122. package/content/.morph/templates/saas/tenant.cs +338 -338
  123. package/content/.morph/templates/service.cs +139 -139
  124. package/content/.morph/templates/sprint-status.yaml +68 -68
  125. package/content/.morph/templates/story.md +143 -143
  126. package/content/.morph/templates/test.cs +239 -239
  127. package/content/.morph/templates/ui-design-system.md +286 -286
  128. package/content/.morph/templates/ui-flows.md +336 -336
  129. package/content/.morph/templates/ui-mockups.md +133 -133
  130. package/content/.morph/test-infra/example.bicep +59 -59
  131. package/content/README.md +79 -79
  132. package/detectors/config-detector.js +223 -223
  133. package/detectors/conversation-analyzer.js +163 -163
  134. package/detectors/index.js +84 -84
  135. package/detectors/standards-generator.js +275 -275
  136. package/docs/api/fonts/Source-Sans-Pro/sourcesanspro-light-webfont.svg +977 -977
  137. package/docs/api/fonts/Source-Sans-Pro/sourcesanspro-regular-webfont.svg +1048 -1048
  138. package/docs/api/scripts/collapse.js +38 -38
  139. package/docs/api/scripts/commonNav.js +28 -28
  140. package/docs/api/scripts/linenumber.js +25 -25
  141. package/docs/api/scripts/nav.js +12 -12
  142. package/docs/api/scripts/polyfill.js +3 -3
  143. package/docs/api/scripts/prettify/Apache-License-2.0.txt +202 -202
  144. package/docs/api/scripts/prettify/lang-css.js +2 -2
  145. package/docs/api/scripts/prettify/prettify.js +28 -28
  146. package/docs/api/scripts/search.js +98 -98
  147. package/docs/api/styles/jsdoc.css +776 -776
  148. package/docs/api/styles/prettify.css +80 -80
  149. package/docs/examples.md +328 -328
  150. package/docs/templates.md +418 -418
  151. package/package.json +1 -1
  152. package/scripts/postinstall.js +132 -132
  153. package/src/commands/advance-phase.js +83 -0
  154. package/src/commands/analyze-blazor-concurrency.js +193 -193
  155. package/src/commands/create-story.js +351 -351
  156. package/src/commands/deploy.js +780 -0
  157. package/src/commands/detect-agents.js +34 -6
  158. package/src/commands/detect.js +104 -104
  159. package/src/commands/generate-context.js +40 -0
  160. package/src/commands/generate.js +149 -149
  161. package/src/commands/lint-fluent.js +352 -352
  162. package/src/commands/rollback-phase.js +185 -185
  163. package/src/commands/session-summary.js +291 -291
  164. package/src/commands/shard-spec.js +224 -224
  165. package/src/commands/sprint-status.js +250 -250
  166. package/src/commands/state.js +333 -333
  167. package/src/commands/sync.js +167 -167
  168. package/src/commands/troubleshoot.js +222 -222
  169. package/src/commands/validate-blazor-state.js +210 -210
  170. package/src/commands/validate-blazor.js +156 -156
  171. package/src/commands/validate-css.js +84 -84
  172. package/src/commands/validate-phase.js +221 -221
  173. package/src/lib/blazor-concurrency-analyzer.js +288 -288
  174. package/src/lib/blazor-state-validator.js +291 -291
  175. package/src/lib/blazor-validator.js +374 -374
  176. package/src/lib/context-generator.js +513 -0
  177. package/src/lib/css-validator.js +352 -352
  178. package/src/lib/design-system-detector.js +187 -0
  179. package/src/lib/design-system-generator.js +298 -298
  180. package/src/lib/design-system-scaffolder.js +299 -0
  181. package/src/lib/hook-executor.js +256 -0
  182. package/src/lib/learning-system.js +520 -520
  183. package/src/lib/mockup-generator.js +366 -366
  184. package/src/lib/spec-validator.js +258 -0
  185. package/src/lib/standards-context-injector.js +287 -0
  186. package/src/lib/team-orchestrator.js +322 -0
  187. package/src/lib/troubleshoot-grep.js +194 -194
  188. package/src/lib/troubleshoot-index.js +144 -144
  189. package/src/lib/ui-detector.js +350 -350
  190. package/src/lib/validation-runner.js +65 -13
  191. package/src/lib/validators/architecture-validator.js +387 -387
  192. package/src/lib/validators/design-system-validator.js +231 -0
  193. package/src/lib/validators/package-validator.js +360 -360
  194. package/src/lib/validators/ui-contrast-validator.js +422 -422
  195. package/src/utils/file-copier.js +9 -1
  196. package/src/utils/logger.js +32 -32
  197. package/src/utils/version-checker.js +175 -175
  198. /package/content/.claude/skills/{checklists → level-0-meta}/code-review.md +0 -0
  199. /package/content/.claude/skills/{checklists → level-0-meta}/simulation-checklist.md +0 -0
  200. /package/content/.claude/skills/{specialists → level-2-domains/ai-agents}/ai-system-architect.md +0 -0
  201. /package/content/.claude/skills/{specialists → level-2-domains/architecture}/po-pm-advisor.md +0 -0
  202. /package/content/.claude/skills/{specialists → level-2-domains/architecture}/standards-architect.md +0 -0
  203. /package/content/.claude/skills/{specialists → level-2-domains/backend}/dotnet-senior.md +0 -0
  204. /package/content/.claude/skills/{specialists → level-2-domains/backend}/ef-modeler.md +0 -0
  205. /package/content/.claude/skills/{specialists → level-2-domains/backend}/hangfire-orchestrator.md +0 -0
  206. /package/content/.claude/skills/{specialists → level-2-domains/backend}/ms-agent-expert.md +0 -0
  207. /package/content/.claude/skills/{stacks/dotnet-blazor.md → level-2-domains/frontend/blazor-builder.md} +0 -0
  208. /package/content/.claude/skills/{stacks/dotnet-nextjs.md → level-2-domains/frontend/nextjs-expert.md} +0 -0
  209. /package/content/.claude/skills/{specialists → level-2-domains/frontend}/ui-ux-designer.md +0 -0
  210. /package/content/.claude/skills/{specialists → level-2-domains/infrastructure}/azure-architect.md +0 -0
  211. /package/content/.claude/skills/{infra → level-2-domains/infrastructure}/bicep-architect.md +0 -0
  212. /package/content/.claude/skills/{infra → level-2-domains/infrastructure}/container-specialist.md +0 -0
  213. /package/content/.claude/skills/{infra → level-2-domains/infrastructure}/devops-engineer.md +0 -0
  214. /package/content/.claude/skills/{integrations → level-2-domains/integrations}/asaas-financial.md +0 -0
  215. /package/content/.claude/skills/{integrations → level-2-domains/integrations}/azure-identity.md +0 -0
  216. /package/content/.claude/skills/{integrations → level-2-domains/integrations}/clerk-auth.md +0 -0
  217. /package/content/.claude/skills/{integrations → level-2-domains/integrations}/resend-email.md +0 -0
  218. /package/content/.claude/skills/{specialists → level-2-domains/quality}/code-analyzer.md +0 -0
@@ -1,536 +1,536 @@
1
- # Vector Search + RAG - EF Core 10 (.NET 10)
2
-
3
- > **Novidade .NET 10:** EF Core 10 suporta vector search nativo no Azure SQL e SQL Server para workloads de AI.
4
-
5
- ---
6
-
7
- ## 🎯 O Que É Vector Search?
8
-
9
- **Vector search** permite buscar dados por **similaridade semântica** ao invés de correspondência exata de texto.
10
-
11
- ### Conceito
12
-
13
- ```
14
- Texto → Embedding (vetor de números) → Busca por similaridade
15
- ```
16
-
17
- **Exemplo:**
18
- - Query: "Como resetar senha?"
19
- - Documento 1: "Tutorial de recuperação de senha" ← **Match semântico!**
20
- - Documento 2: "Alterar credenciais de acesso" ← **Match semântico!**
21
- - Documento 3: "Configurar email" ← Não match
22
-
23
- ### Casos de Uso
24
-
25
- | Caso de Uso | Descrição |
26
- |-------------|-----------|
27
- | **RAG (Retrieval-Augmented Generation)** | Buscar documentos relevantes para enviar ao LLM |
28
- | **Busca semântica** | Encontrar conteúdo similar sem keywords exatas |
29
- | **Recomendações** | Sugerir produtos/artigos similares |
30
- | **Deduplicação** | Identificar conteúdo duplicado |
31
-
32
- ---
33
-
34
- ## 📦 Setup
35
-
36
- ### 1. Packages Necessários
37
-
38
- ```xml
39
- <!-- .csproj -->
40
- <PackageReference Include="Microsoft.EntityFrameworkCore.SqlServer" Version="10.0.0" />
41
- <PackageReference Include="Microsoft.Extensions.AI.Embeddings" Version="1.0.0" />
42
- ```
43
-
44
- ### 2. Azure SQL / SQL Server Requerido
45
-
46
- **Requisito:** Azure SQL Database ou SQL Server 2022+
47
-
48
- **Nota:** Vector search não funciona em LocalDB.
49
-
50
- ---
51
-
52
- ## 🗄️ Modelo de Dados com Vectors
53
-
54
- ### Entidade com Embedding
55
-
56
- ```csharp
57
- using Microsoft.EntityFrameworkCore;
58
- using System.Numerics.Tensors;
59
-
60
- public class Document
61
- {
62
- public int Id { get; set; }
63
- public string Title { get; set; } = null!;
64
- public string Content { get; set; } = null!;
65
- public DateTime CreatedAt { get; set; } = DateTime.UtcNow;
66
-
67
- // Vector embedding (1536 dimensões para text-embedding-3-small)
68
- public ReadOnlyMemory<float> Embedding { get; set; }
69
- }
70
- ```
71
-
72
- ### DbContext Configuração
73
-
74
- ```csharp
75
- public class AppDbContext : DbContext
76
- {
77
- public DbSet<Document> Documents { get; set; }
78
-
79
- protected override void OnModelCreating(ModelBuilder modelBuilder)
80
- {
81
- modelBuilder.Entity<Document>(entity =>
82
- {
83
- entity.HasKey(d => d.Id);
84
-
85
- // Configurar coluna de vector
86
- entity.Property(d => d.Embedding)
87
- .HasColumnType("vector(1536)") // 1536 = dimensões do embedding
88
- .IsRequired();
89
-
90
- // Índice de vector para performance
91
- entity.HasIndex(d => d.Embedding)
92
- .HasMethod("ivfflat") // ou "hnsw"
93
- .HasOptions("lists = 100");
94
- });
95
- }
96
- }
97
- ```
98
-
99
- ### Migration
100
-
101
- ```bash
102
- dotnet ef migrations add AddVectorSearch
103
- dotnet ef database update
104
- ```
105
-
106
- **SQL Gerado:**
107
- ```sql
108
- ALTER TABLE Documents ADD Embedding vector(1536) NOT NULL;
109
- CREATE INDEX IX_Documents_Embedding ON Documents USING ivfflat (Embedding) WITH (lists = 100);
110
- ```
111
-
112
- ---
113
-
114
- ## 🔢 Gerando Embeddings
115
-
116
- ### Serviço de Embedding
117
-
118
- ```csharp
119
- using Microsoft.Extensions.AI;
120
-
121
- public interface IEmbeddingService
122
- {
123
- Task<ReadOnlyMemory<float>> GenerateEmbeddingAsync(string text, CancellationToken ct = default);
124
- }
125
-
126
- public class EmbeddingService : IEmbeddingService
127
- {
128
- private readonly IEmbeddingGenerator<string, Embedding<float>> _embeddingGenerator;
129
-
130
- public EmbeddingService(IEmbeddingGenerator<string, Embedding<float>> embeddingGenerator)
131
- {
132
- _embeddingGenerator = embeddingGenerator;
133
- }
134
-
135
- public async Task<ReadOnlyMemory<float>> GenerateEmbeddingAsync(
136
- string text,
137
- CancellationToken ct = default)
138
- {
139
- var embeddings = await _embeddingGenerator.GenerateAsync([text], cancellationToken: ct);
140
- return embeddings[0].Vector;
141
- }
142
- }
143
- ```
144
-
145
- ### Configuração no Program.cs
146
-
147
- ```csharp
148
- using Microsoft.Extensions.AI;
149
-
150
- builder.Services.AddSingleton<IEmbeddingGenerator<string, Embedding<float>>>(sp =>
151
- {
152
- var config = sp.GetRequiredService<IConfiguration>();
153
-
154
- return new EmbeddingGenerator(
155
- model: "text-embedding-3-small", // 1536 dimensões
156
- credential: new ApiKeyCredential(config["AzureOpenAI:ApiKey"]!),
157
- endpoint: new Uri(config["AzureOpenAI:Endpoint"]!)
158
- );
159
- });
160
-
161
- builder.Services.AddScoped<IEmbeddingService, EmbeddingService>();
162
- ```
163
-
164
- ---
165
-
166
- ## 🔍 Vector Search com EF Core 10
167
-
168
- ### Query de Similaridade
169
-
170
- ```csharp
171
- using Microsoft.EntityFrameworkCore;
172
-
173
- public class DocumentSearchService
174
- {
175
- private readonly AppDbContext _context;
176
- private readonly IEmbeddingService _embeddingService;
177
-
178
- public DocumentSearchService(
179
- AppDbContext context,
180
- IEmbeddingService embeddingService)
181
- {
182
- _context = context;
183
- _embeddingService = embeddingService;
184
- }
185
-
186
- public async Task<List<DocumentSearchResult>> SearchAsync(
187
- string query,
188
- int limit = 5,
189
- CancellationToken ct = default)
190
- {
191
- // 1. Gerar embedding da query
192
- var queryEmbedding = await _embeddingService.GenerateEmbeddingAsync(query, ct);
193
-
194
- // 2. Buscar documentos similares
195
- var results = await _context.Documents
196
- .Select(d => new DocumentSearchResult
197
- {
198
- Document = d,
199
- // Calcular distância (menor = mais similar)
200
- Distance = EF.Functions.VectorDistance(d.Embedding, queryEmbedding, DistanceFunction.Cosine)
201
- })
202
- .OrderBy(r => r.Distance)
203
- .Take(limit)
204
- .ToListAsync(ct);
205
-
206
- return results;
207
- }
208
- }
209
-
210
- public class DocumentSearchResult
211
- {
212
- public Document Document { get; set; } = null!;
213
- public double Distance { get; set; }
214
- public double Similarity => 1 - Distance; // Converter distância em similaridade
215
- }
216
- ```
217
-
218
- ### Funções de Distância
219
-
220
- | Função | Descrição | Quando Usar |
221
- |--------|-----------|-------------|
222
- | `Cosine` | Distância cosseno (0-2) | **Recomendado** para texto |
223
- | `Euclidean` | Distância euclidiana | Dados numéricos |
224
- | `DotProduct` | Produto escalar | Vetores normalizados |
225
-
226
- ```csharp
227
- EF.Functions.VectorDistance(vector1, vector2, DistanceFunction.Cosine)
228
- ```
229
-
230
- ---
231
-
232
- ## 🤖 RAG Pattern Completo
233
-
234
- ### Implementação RAG com Agent Framework
235
-
236
- ```csharp
237
- using Microsoft.Agents.AI;
238
-
239
- public interface IDocumentAssistantAgent
240
- {
241
- Task<string> AskQuestionAsync(string question, CancellationToken ct = default);
242
- }
243
-
244
- public class DocumentAssistantAgent : IDocumentAssistantAgent
245
- {
246
- private readonly IChatClient _chatClient;
247
- private readonly DocumentSearchService _searchService;
248
- private readonly ILogger<DocumentAssistantAgent> _logger;
249
-
250
- public DocumentAssistantAgent(
251
- IChatClient chatClient,
252
- DocumentSearchService searchService,
253
- ILogger<DocumentAssistantAgent> logger)
254
- {
255
- _chatClient = chatClient;
256
- _searchService = searchService;
257
- _logger = logger;
258
- }
259
-
260
- public async Task<string> AskQuestionAsync(string question, CancellationToken ct = default)
261
- {
262
- // 1. Retrieval: Buscar documentos relevantes
263
- var relevantDocs = await _searchService.SearchAsync(question, limit: 3, ct);
264
-
265
- _logger.LogInformation(
266
- "Encontrados {Count} documentos relevantes para: {Question}",
267
- relevantDocs.Count,
268
- question
269
- );
270
-
271
- // 2. Construir contexto
272
- var context = string.Join("\n\n", relevantDocs.Select(r =>
273
- $"[Documento {r.Document.Id} - Similaridade: {r.Similarity:P0}]\n" +
274
- $"Título: {r.Document.Title}\n" +
275
- $"Conteúdo: {r.Document.Content}"
276
- ));
277
-
278
- // 3. Augmentation: Criar agente com contexto
279
- var agent = _chatClient.CreateAgent(
280
- instructions: """
281
- Você é um assistente que responde perguntas baseado em documentos fornecidos.
282
-
283
- Regras:
284
- 1. Use APENAS informações dos documentos fornecidos
285
- 2. Se a informação não estiver nos documentos, diga "Não encontrei informação sobre isso"
286
- 3. Cite o número do documento ao responder
287
- 4. Seja conciso e objetivo
288
-
289
- Documentos disponíveis:
290
- """ + context,
291
- name: "DocumentAssistant"
292
- );
293
-
294
- // 4. Generation: Gerar resposta
295
- var response = await agent.RunAsync(question, cancellationToken: ct);
296
-
297
- return response.Content;
298
- }
299
- }
300
- ```
301
-
302
- ### Uso no Blazor
303
-
304
- ```razor
305
- @page "/ask"
306
- @inject IDocumentAssistantAgent Assistant
307
-
308
- <h3>Assistente de Documentos</h3>
309
-
310
- <EditForm Model="_input" OnValidSubmit="AskQuestion">
311
- <InputText @bind-Value="_input.Question" placeholder="Faça uma pergunta..." />
312
- <button type="submit" disabled="@_isLoading">Perguntar</button>
313
- </EditForm>
314
-
315
- @if (_isLoading)
316
- {
317
- <p>Buscando resposta...</p>
318
- }
319
- else if (!string.IsNullOrEmpty(_answer))
320
- {
321
- <div class="answer">
322
- <strong>Resposta:</strong>
323
- <p>@_answer</p>
324
- </div>
325
- }
326
-
327
- @code {
328
- private QuestionInput _input = new();
329
- private string _answer = "";
330
- private bool _isLoading;
331
-
332
- private async Task AskQuestion()
333
- {
334
- _isLoading = true;
335
- _answer = "";
336
-
337
- try
338
- {
339
- _answer = await Assistant.AskQuestionAsync(_input.Question);
340
- }
341
- finally
342
- {
343
- _isLoading = false;
344
- }
345
- }
346
-
347
- public class QuestionInput
348
- {
349
- public string Question { get; set; } = "";
350
- }
351
- }
352
- ```
353
-
354
- ---
355
-
356
- ## 📥 Indexação de Documentos
357
-
358
- ### Serviço de Indexação
359
-
360
- ```csharp
361
- public class DocumentIndexingService
362
- {
363
- private readonly AppDbContext _context;
364
- private readonly IEmbeddingService _embeddingService;
365
-
366
- public async Task IndexDocumentAsync(
367
- string title,
368
- string content,
369
- CancellationToken ct = default)
370
- {
371
- // 1. Gerar embedding do conteúdo
372
- var embedding = await _embeddingService.GenerateEmbeddingAsync(
373
- $"{title}\n{content}", // Combinar título e conteúdo
374
- ct
375
- );
376
-
377
- // 2. Criar documento
378
- var document = new Document
379
- {
380
- Title = title,
381
- Content = content,
382
- Embedding = embedding,
383
- CreatedAt = DateTime.UtcNow
384
- };
385
-
386
- // 3. Salvar no banco
387
- _context.Documents.Add(document);
388
- await _context.SaveChangesAsync(ct);
389
- }
390
-
391
- public async Task BulkIndexAsync(
392
- List<(string Title, string Content)> documents,
393
- CancellationToken ct = default)
394
- {
395
- foreach (var (title, content) in documents)
396
- {
397
- await IndexDocumentAsync(title, content, ct);
398
- }
399
- }
400
- }
401
- ```
402
-
403
- ### Job de Indexação com Hangfire
404
-
405
- ```csharp
406
- public class DocumentIndexingJob
407
- {
408
- private readonly DocumentIndexingService _indexingService;
409
- private readonly IDocumentProvider _documentProvider;
410
-
411
- public async Task IndexAllDocumentsAsync()
412
- {
413
- // Buscar documentos de fonte externa (API, arquivos, etc.)
414
- var documents = await _documentProvider.GetAllDocumentsAsync();
415
-
416
- await _indexingService.BulkIndexAsync(documents);
417
- }
418
- }
419
-
420
- // Program.cs - Agendar job diário
421
- RecurringJob.AddOrUpdate<DocumentIndexingJob>(
422
- "index-documents",
423
- job => job.IndexAllDocumentsAsync(),
424
- Cron.Daily
425
- );
426
- ```
427
-
428
- ---
429
-
430
- ## 📊 Índices de Performance
431
-
432
- ### Tipos de Índices
433
-
434
- | Tipo | Descrição | Performance | Precisão |
435
- |------|-----------|-------------|----------|
436
- | **IVFFlat** | Inverted File + Flat compression | Boa | Alta |
437
- | **HNSW** | Hierarchical Navigable Small World | Excelente | Alta |
438
-
439
- ### Configuração IVFFlat
440
-
441
- ```csharp
442
- entity.HasIndex(d => d.Embedding)
443
- .HasMethod("ivfflat")
444
- .HasOptions("lists = 100"); // Ajustar conforme dataset
445
- ```
446
-
447
- **Recomendação de `lists`:**
448
- - Pequeno dataset (<10k docs): `lists = 50`
449
- - Médio dataset (10k-100k): `lists = 100`
450
- - Grande dataset (>100k): `lists = 500+`
451
-
452
- ### Configuração HNSW
453
-
454
- ```csharp
455
- entity.HasIndex(d => d.Embedding)
456
- .HasMethod("hnsw")
457
- .HasOptions("m = 16, ef_construction = 64");
458
- ```
459
-
460
- **Parâmetros:**
461
- - `m`: Número de conexões (padrão: 16)
462
- - `ef_construction`: Qualidade do índice (padrão: 64)
463
-
464
- ---
465
-
466
- ## 💰 Custos
467
-
468
- ### Embedding Generation
469
-
470
- | Modelo | Dimensões | Custo |
471
- |--------|-----------|-------|
472
- | text-embedding-3-small | 1536 | $0.02 / 1M tokens |
473
- | text-embedding-3-large | 3072 | $0.13 / 1M tokens |
474
-
475
- **Recomendação:** Use `text-embedding-3-small` (melhor custo-benefício).
476
-
477
- ### Storage
478
-
479
- | Dimensões | Tamanho por Documento | 10k Docs | 100k Docs |
480
- |-----------|-----------------------|----------|-----------|
481
- | 1536 | ~6 KB | ~60 MB | ~600 MB |
482
- | 3072 | ~12 KB | ~120 MB | ~1.2 GB |
483
-
484
- ---
485
-
486
- ## ✅ Checklist de Implementação
487
-
488
- - [ ] EF Core 10 instalado
489
- - [ ] Azure SQL ou SQL Server 2022+
490
- - [ ] Entidade com propriedade `ReadOnlyMemory<float>`
491
- - [ ] Índice de vector criado (`ivfflat` ou `hnsw`)
492
- - [ ] `IEmbeddingService` configurado
493
- - [ ] Serviço de search implementado
494
- - [ ] RAG pattern com Agent Framework
495
- - [ ] Job de indexação configurado
496
- - [ ] Testes de similaridade funcionando
497
-
498
- ---
499
-
500
- ## 🐛 Troubleshooting
501
-
502
- ### Erro: "vector type not supported"
503
-
504
- **Causa:** SQL Server não suporta vectors ou versão antiga.
505
-
506
- **Solução:** Use Azure SQL Database ou SQL Server 2022+.
507
-
508
- ### Performance lenta em queries
509
-
510
- **Causa:** Índice não criado ou mal configurado.
511
-
512
- **Solução:**
513
- - Verifique se índice foi criado: `SELECT * FROM sys.indexes WHERE name LIKE '%Embedding%'`
514
- - Ajuste parâmetros `lists` (IVFFlat) ou `m` (HNSW)
515
-
516
- ### Embeddings com dimensões erradas
517
-
518
- **Causa:** Modelo de embedding diferente do esperado.
519
-
520
- **Solução:**
521
- - `text-embedding-3-small` → 1536 dimensões
522
- - `text-embedding-3-large` → 3072 dimensões
523
- - Atualize `vector(N)` na migration
524
-
525
- ---
526
-
527
- ## 📚 Referências
528
-
529
- - [EF Core 10 - Vector Search](https://learn.microsoft.com/ef/core/what-is-new/ef-core-10.0/whatsnew)
530
- - [Azure SQL Vector Search](https://learn.microsoft.com/azure/azure-sql/database/ai-artificial-intelligence-vector-search)
531
- - [OpenAI Embeddings](https://platform.openai.com/docs/guides/embeddings)
532
- - [RAG Architecture](https://learn.microsoft.com/azure/architecture/ai-ml/guide/rag/rag-solution-design-and-evaluation-guide)
533
-
534
- ---
535
-
536
- *MORPH-SPEC by Polymorphism Tech*
1
+ # Vector Search + RAG - EF Core 10 (.NET 10)
2
+
3
+ > **Novidade .NET 10:** EF Core 10 suporta vector search nativo no Azure SQL e SQL Server para workloads de AI.
4
+
5
+ ---
6
+
7
+ ## 🎯 O Que É Vector Search?
8
+
9
+ **Vector search** permite buscar dados por **similaridade semântica** ao invés de correspondência exata de texto.
10
+
11
+ ### Conceito
12
+
13
+ ```
14
+ Texto → Embedding (vetor de números) → Busca por similaridade
15
+ ```
16
+
17
+ **Exemplo:**
18
+ - Query: "Como resetar senha?"
19
+ - Documento 1: "Tutorial de recuperação de senha" ← **Match semântico!**
20
+ - Documento 2: "Alterar credenciais de acesso" ← **Match semântico!**
21
+ - Documento 3: "Configurar email" ← Não match
22
+
23
+ ### Casos de Uso
24
+
25
+ | Caso de Uso | Descrição |
26
+ |-------------|-----------|
27
+ | **RAG (Retrieval-Augmented Generation)** | Buscar documentos relevantes para enviar ao LLM |
28
+ | **Busca semântica** | Encontrar conteúdo similar sem keywords exatas |
29
+ | **Recomendações** | Sugerir produtos/artigos similares |
30
+ | **Deduplicação** | Identificar conteúdo duplicado |
31
+
32
+ ---
33
+
34
+ ## 📦 Setup
35
+
36
+ ### 1. Packages Necessários
37
+
38
+ ```xml
39
+ <!-- .csproj -->
40
+ <PackageReference Include="Microsoft.EntityFrameworkCore.SqlServer" Version="10.0.0" />
41
+ <PackageReference Include="Microsoft.Extensions.AI.Embeddings" Version="1.0.0" />
42
+ ```
43
+
44
+ ### 2. Azure SQL / SQL Server Requerido
45
+
46
+ **Requisito:** Azure SQL Database ou SQL Server 2022+
47
+
48
+ **Nota:** Vector search não funciona em LocalDB.
49
+
50
+ ---
51
+
52
+ ## 🗄️ Modelo de Dados com Vectors
53
+
54
+ ### Entidade com Embedding
55
+
56
+ ```csharp
57
+ using Microsoft.EntityFrameworkCore;
58
+ using System.Numerics.Tensors;
59
+
60
+ public class Document
61
+ {
62
+ public int Id { get; set; }
63
+ public string Title { get; set; } = null!;
64
+ public string Content { get; set; } = null!;
65
+ public DateTime CreatedAt { get; set; } = DateTime.UtcNow;
66
+
67
+ // Vector embedding (1536 dimensões para text-embedding-3-small)
68
+ public ReadOnlyMemory<float> Embedding { get; set; }
69
+ }
70
+ ```
71
+
72
+ ### DbContext Configuração
73
+
74
+ ```csharp
75
+ public class AppDbContext : DbContext
76
+ {
77
+ public DbSet<Document> Documents { get; set; }
78
+
79
+ protected override void OnModelCreating(ModelBuilder modelBuilder)
80
+ {
81
+ modelBuilder.Entity<Document>(entity =>
82
+ {
83
+ entity.HasKey(d => d.Id);
84
+
85
+ // Configurar coluna de vector
86
+ entity.Property(d => d.Embedding)
87
+ .HasColumnType("vector(1536)") // 1536 = dimensões do embedding
88
+ .IsRequired();
89
+
90
+ // Índice de vector para performance
91
+ entity.HasIndex(d => d.Embedding)
92
+ .HasMethod("ivfflat") // ou "hnsw"
93
+ .HasOptions("lists = 100");
94
+ });
95
+ }
96
+ }
97
+ ```
98
+
99
+ ### Migration
100
+
101
+ ```bash
102
+ dotnet ef migrations add AddVectorSearch
103
+ dotnet ef database update
104
+ ```
105
+
106
+ **SQL Gerado:**
107
+ ```sql
108
+ ALTER TABLE Documents ADD Embedding vector(1536) NOT NULL;
109
+ CREATE INDEX IX_Documents_Embedding ON Documents USING ivfflat (Embedding) WITH (lists = 100);
110
+ ```
111
+
112
+ ---
113
+
114
+ ## 🔢 Gerando Embeddings
115
+
116
+ ### Serviço de Embedding
117
+
118
+ ```csharp
119
+ using Microsoft.Extensions.AI;
120
+
121
+ public interface IEmbeddingService
122
+ {
123
+ Task<ReadOnlyMemory<float>> GenerateEmbeddingAsync(string text, CancellationToken ct = default);
124
+ }
125
+
126
+ public class EmbeddingService : IEmbeddingService
127
+ {
128
+ private readonly IEmbeddingGenerator<string, Embedding<float>> _embeddingGenerator;
129
+
130
+ public EmbeddingService(IEmbeddingGenerator<string, Embedding<float>> embeddingGenerator)
131
+ {
132
+ _embeddingGenerator = embeddingGenerator;
133
+ }
134
+
135
+ public async Task<ReadOnlyMemory<float>> GenerateEmbeddingAsync(
136
+ string text,
137
+ CancellationToken ct = default)
138
+ {
139
+ var embeddings = await _embeddingGenerator.GenerateAsync([text], cancellationToken: ct);
140
+ return embeddings[0].Vector;
141
+ }
142
+ }
143
+ ```
144
+
145
+ ### Configuração no Program.cs
146
+
147
+ ```csharp
148
+ using Microsoft.Extensions.AI;
149
+
150
+ builder.Services.AddSingleton<IEmbeddingGenerator<string, Embedding<float>>>(sp =>
151
+ {
152
+ var config = sp.GetRequiredService<IConfiguration>();
153
+
154
+ return new EmbeddingGenerator(
155
+ model: "text-embedding-3-small", // 1536 dimensões
156
+ credential: new ApiKeyCredential(config["AzureOpenAI:ApiKey"]!),
157
+ endpoint: new Uri(config["AzureOpenAI:Endpoint"]!)
158
+ );
159
+ });
160
+
161
+ builder.Services.AddScoped<IEmbeddingService, EmbeddingService>();
162
+ ```
163
+
164
+ ---
165
+
166
+ ## 🔍 Vector Search com EF Core 10
167
+
168
+ ### Query de Similaridade
169
+
170
+ ```csharp
171
+ using Microsoft.EntityFrameworkCore;
172
+
173
+ public class DocumentSearchService
174
+ {
175
+ private readonly AppDbContext _context;
176
+ private readonly IEmbeddingService _embeddingService;
177
+
178
+ public DocumentSearchService(
179
+ AppDbContext context,
180
+ IEmbeddingService embeddingService)
181
+ {
182
+ _context = context;
183
+ _embeddingService = embeddingService;
184
+ }
185
+
186
+ public async Task<List<DocumentSearchResult>> SearchAsync(
187
+ string query,
188
+ int limit = 5,
189
+ CancellationToken ct = default)
190
+ {
191
+ // 1. Gerar embedding da query
192
+ var queryEmbedding = await _embeddingService.GenerateEmbeddingAsync(query, ct);
193
+
194
+ // 2. Buscar documentos similares
195
+ var results = await _context.Documents
196
+ .Select(d => new DocumentSearchResult
197
+ {
198
+ Document = d,
199
+ // Calcular distância (menor = mais similar)
200
+ Distance = EF.Functions.VectorDistance(d.Embedding, queryEmbedding, DistanceFunction.Cosine)
201
+ })
202
+ .OrderBy(r => r.Distance)
203
+ .Take(limit)
204
+ .ToListAsync(ct);
205
+
206
+ return results;
207
+ }
208
+ }
209
+
210
+ public class DocumentSearchResult
211
+ {
212
+ public Document Document { get; set; } = null!;
213
+ public double Distance { get; set; }
214
+ public double Similarity => 1 - Distance; // Converter distância em similaridade
215
+ }
216
+ ```
217
+
218
+ ### Funções de Distância
219
+
220
+ | Função | Descrição | Quando Usar |
221
+ |--------|-----------|-------------|
222
+ | `Cosine` | Distância cosseno (0-2) | **Recomendado** para texto |
223
+ | `Euclidean` | Distância euclidiana | Dados numéricos |
224
+ | `DotProduct` | Produto escalar | Vetores normalizados |
225
+
226
+ ```csharp
227
+ EF.Functions.VectorDistance(vector1, vector2, DistanceFunction.Cosine)
228
+ ```
229
+
230
+ ---
231
+
232
+ ## 🤖 RAG Pattern Completo
233
+
234
+ ### Implementação RAG com Agent Framework
235
+
236
+ ```csharp
237
+ using Microsoft.Agents.AI;
238
+
239
+ public interface IDocumentAssistantAgent
240
+ {
241
+ Task<string> AskQuestionAsync(string question, CancellationToken ct = default);
242
+ }
243
+
244
+ public class DocumentAssistantAgent : IDocumentAssistantAgent
245
+ {
246
+ private readonly IChatClient _chatClient;
247
+ private readonly DocumentSearchService _searchService;
248
+ private readonly ILogger<DocumentAssistantAgent> _logger;
249
+
250
+ public DocumentAssistantAgent(
251
+ IChatClient chatClient,
252
+ DocumentSearchService searchService,
253
+ ILogger<DocumentAssistantAgent> logger)
254
+ {
255
+ _chatClient = chatClient;
256
+ _searchService = searchService;
257
+ _logger = logger;
258
+ }
259
+
260
+ public async Task<string> AskQuestionAsync(string question, CancellationToken ct = default)
261
+ {
262
+ // 1. Retrieval: Buscar documentos relevantes
263
+ var relevantDocs = await _searchService.SearchAsync(question, limit: 3, ct);
264
+
265
+ _logger.LogInformation(
266
+ "Encontrados {Count} documentos relevantes para: {Question}",
267
+ relevantDocs.Count,
268
+ question
269
+ );
270
+
271
+ // 2. Construir contexto
272
+ var context = string.Join("\n\n", relevantDocs.Select(r =>
273
+ $"[Documento {r.Document.Id} - Similaridade: {r.Similarity:P0}]\n" +
274
+ $"Título: {r.Document.Title}\n" +
275
+ $"Conteúdo: {r.Document.Content}"
276
+ ));
277
+
278
+ // 3. Augmentation: Criar agente com contexto
279
+ var agent = _chatClient.CreateAgent(
280
+ instructions: """
281
+ Você é um assistente que responde perguntas baseado em documentos fornecidos.
282
+
283
+ Regras:
284
+ 1. Use APENAS informações dos documentos fornecidos
285
+ 2. Se a informação não estiver nos documentos, diga "Não encontrei informação sobre isso"
286
+ 3. Cite o número do documento ao responder
287
+ 4. Seja conciso e objetivo
288
+
289
+ Documentos disponíveis:
290
+ """ + context,
291
+ name: "DocumentAssistant"
292
+ );
293
+
294
+ // 4. Generation: Gerar resposta
295
+ var response = await agent.RunAsync(question, cancellationToken: ct);
296
+
297
+ return response.Content;
298
+ }
299
+ }
300
+ ```
301
+
302
+ ### Uso no Blazor
303
+
304
+ ```razor
305
+ @page "/ask"
306
+ @inject IDocumentAssistantAgent Assistant
307
+
308
+ <h3>Assistente de Documentos</h3>
309
+
310
+ <EditForm Model="_input" OnValidSubmit="AskQuestion">
311
+ <InputText @bind-Value="_input.Question" placeholder="Faça uma pergunta..." />
312
+ <button type="submit" disabled="@_isLoading">Perguntar</button>
313
+ </EditForm>
314
+
315
+ @if (_isLoading)
316
+ {
317
+ <p>Buscando resposta...</p>
318
+ }
319
+ else if (!string.IsNullOrEmpty(_answer))
320
+ {
321
+ <div class="answer">
322
+ <strong>Resposta:</strong>
323
+ <p>@_answer</p>
324
+ </div>
325
+ }
326
+
327
+ @code {
328
+ private QuestionInput _input = new();
329
+ private string _answer = "";
330
+ private bool _isLoading;
331
+
332
+ private async Task AskQuestion()
333
+ {
334
+ _isLoading = true;
335
+ _answer = "";
336
+
337
+ try
338
+ {
339
+ _answer = await Assistant.AskQuestionAsync(_input.Question);
340
+ }
341
+ finally
342
+ {
343
+ _isLoading = false;
344
+ }
345
+ }
346
+
347
+ public class QuestionInput
348
+ {
349
+ public string Question { get; set; } = "";
350
+ }
351
+ }
352
+ ```
353
+
354
+ ---
355
+
356
+ ## 📥 Indexação de Documentos
357
+
358
+ ### Serviço de Indexação
359
+
360
+ ```csharp
361
+ public class DocumentIndexingService
362
+ {
363
+ private readonly AppDbContext _context;
364
+ private readonly IEmbeddingService _embeddingService;
365
+
366
+ public async Task IndexDocumentAsync(
367
+ string title,
368
+ string content,
369
+ CancellationToken ct = default)
370
+ {
371
+ // 1. Gerar embedding do conteúdo
372
+ var embedding = await _embeddingService.GenerateEmbeddingAsync(
373
+ $"{title}\n{content}", // Combinar título e conteúdo
374
+ ct
375
+ );
376
+
377
+ // 2. Criar documento
378
+ var document = new Document
379
+ {
380
+ Title = title,
381
+ Content = content,
382
+ Embedding = embedding,
383
+ CreatedAt = DateTime.UtcNow
384
+ };
385
+
386
+ // 3. Salvar no banco
387
+ _context.Documents.Add(document);
388
+ await _context.SaveChangesAsync(ct);
389
+ }
390
+
391
+ public async Task BulkIndexAsync(
392
+ List<(string Title, string Content)> documents,
393
+ CancellationToken ct = default)
394
+ {
395
+ foreach (var (title, content) in documents)
396
+ {
397
+ await IndexDocumentAsync(title, content, ct);
398
+ }
399
+ }
400
+ }
401
+ ```
402
+
403
+ ### Job de Indexação com Hangfire
404
+
405
+ ```csharp
406
+ public class DocumentIndexingJob
407
+ {
408
+ private readonly DocumentIndexingService _indexingService;
409
+ private readonly IDocumentProvider _documentProvider;
410
+
411
+ public async Task IndexAllDocumentsAsync()
412
+ {
413
+ // Buscar documentos de fonte externa (API, arquivos, etc.)
414
+ var documents = await _documentProvider.GetAllDocumentsAsync();
415
+
416
+ await _indexingService.BulkIndexAsync(documents);
417
+ }
418
+ }
419
+
420
+ // Program.cs - Agendar job diário
421
+ RecurringJob.AddOrUpdate<DocumentIndexingJob>(
422
+ "index-documents",
423
+ job => job.IndexAllDocumentsAsync(),
424
+ Cron.Daily
425
+ );
426
+ ```
427
+
428
+ ---
429
+
430
+ ## 📊 Índices de Performance
431
+
432
+ ### Tipos de Índices
433
+
434
+ | Tipo | Descrição | Performance | Precisão |
435
+ |------|-----------|-------------|----------|
436
+ | **IVFFlat** | Inverted File + Flat compression | Boa | Alta |
437
+ | **HNSW** | Hierarchical Navigable Small World | Excelente | Alta |
438
+
439
+ ### Configuração IVFFlat
440
+
441
+ ```csharp
442
+ entity.HasIndex(d => d.Embedding)
443
+ .HasMethod("ivfflat")
444
+ .HasOptions("lists = 100"); // Ajustar conforme dataset
445
+ ```
446
+
447
+ **Recomendação de `lists`:**
448
+ - Pequeno dataset (<10k docs): `lists = 50`
449
+ - Médio dataset (10k-100k): `lists = 100`
450
+ - Grande dataset (>100k): `lists = 500+`
451
+
452
+ ### Configuração HNSW
453
+
454
+ ```csharp
455
+ entity.HasIndex(d => d.Embedding)
456
+ .HasMethod("hnsw")
457
+ .HasOptions("m = 16, ef_construction = 64");
458
+ ```
459
+
460
+ **Parâmetros:**
461
+ - `m`: Número de conexões (padrão: 16)
462
+ - `ef_construction`: Qualidade do índice (padrão: 64)
463
+
464
+ ---
465
+
466
+ ## 💰 Custos
467
+
468
+ ### Embedding Generation
469
+
470
+ | Modelo | Dimensões | Custo |
471
+ |--------|-----------|-------|
472
+ | text-embedding-3-small | 1536 | $0.02 / 1M tokens |
473
+ | text-embedding-3-large | 3072 | $0.13 / 1M tokens |
474
+
475
+ **Recomendação:** Use `text-embedding-3-small` (melhor custo-benefício).
476
+
477
+ ### Storage
478
+
479
+ | Dimensões | Tamanho por Documento | 10k Docs | 100k Docs |
480
+ |-----------|-----------------------|----------|-----------|
481
+ | 1536 | ~6 KB | ~60 MB | ~600 MB |
482
+ | 3072 | ~12 KB | ~120 MB | ~1.2 GB |
483
+
484
+ ---
485
+
486
+ ## ✅ Checklist de Implementação
487
+
488
+ - [ ] EF Core 10 instalado
489
+ - [ ] Azure SQL ou SQL Server 2022+
490
+ - [ ] Entidade com propriedade `ReadOnlyMemory<float>`
491
+ - [ ] Índice de vector criado (`ivfflat` ou `hnsw`)
492
+ - [ ] `IEmbeddingService` configurado
493
+ - [ ] Serviço de search implementado
494
+ - [ ] RAG pattern com Agent Framework
495
+ - [ ] Job de indexação configurado
496
+ - [ ] Testes de similaridade funcionando
497
+
498
+ ---
499
+
500
+ ## 🐛 Troubleshooting
501
+
502
+ ### Erro: "vector type not supported"
503
+
504
+ **Causa:** SQL Server não suporta vectors ou versão antiga.
505
+
506
+ **Solução:** Use Azure SQL Database ou SQL Server 2022+.
507
+
508
+ ### Performance lenta em queries
509
+
510
+ **Causa:** Índice não criado ou mal configurado.
511
+
512
+ **Solução:**
513
+ - Verifique se índice foi criado: `SELECT * FROM sys.indexes WHERE name LIKE '%Embedding%'`
514
+ - Ajuste parâmetros `lists` (IVFFlat) ou `m` (HNSW)
515
+
516
+ ### Embeddings com dimensões erradas
517
+
518
+ **Causa:** Modelo de embedding diferente do esperado.
519
+
520
+ **Solução:**
521
+ - `text-embedding-3-small` → 1536 dimensões
522
+ - `text-embedding-3-large` → 3072 dimensões
523
+ - Atualize `vector(N)` na migration
524
+
525
+ ---
526
+
527
+ ## 📚 Referências
528
+
529
+ - [EF Core 10 - Vector Search](https://learn.microsoft.com/ef/core/what-is-new/ef-core-10.0/whatsnew)
530
+ - [Azure SQL Vector Search](https://learn.microsoft.com/azure/azure-sql/database/ai-artificial-intelligence-vector-search)
531
+ - [OpenAI Embeddings](https://platform.openai.com/docs/guides/embeddings)
532
+ - [RAG Architecture](https://learn.microsoft.com/azure/architecture/ai-ml/guide/rag/rag-solution-design-and-evaluation-guide)
533
+
534
+ ---
535
+
536
+ *MORPH-SPEC by Polymorphism Tech*