End-to-End Workflows Test Suite

Workflows completos ponta a ponta devem funcionar corretamente, desde inicialização até resultado final, passando por múltiplos componentes e integrações sem falhas. Esta suite valida cenários reais de usuário através de toda a stack. Cobertura: 100+ testes | Prioridade: ALTA

Gemini CLI Complete Flow

• Initialize CLI + auth (8 testes)
• Query Vectora through Gemini (10 testes)
• Get multi-turn conversation (8 testes)
• Context persistence (8 testes)
• Error recovery (5 testes)

SLA: Complete query response in < 3s

Gemini CLI Scenario

User: "How do I validate JWT tokens?"
→ CLI initializes with auth
→ Gemini CLI calls Vectora search
→ Vectora returns code examples
→ Gemini formats response
→ User sees answer in < 2s

VS Code Extension Workflow

• Open extension in VS Code (8 testes)
• Select code + search (10 testes)
• View results inline (8 testes)
• Apply suggestions (8 testes)
• Multi-file workflow (8 testes)

SLA: Results shown in < 500ms from selection

VS Code Extension Scenario

User: Selects "getUserById" in editor
→ Right-click "Search with Vectora"
→ Vectora searches for references
→ Results shown in sidebar
→ User sees 10+ usages in < 200ms

MCP Protocol Flow

• Client connects to server (8 testes)
• Lists available tools (5 testes)
• Invokes tool with params (10 testes)
• Receives result correctly (8 testes)
• Handles errors gracefully (8 testes)

SLA: Tool invocation response in < 1s

MCP Protocol Scenario

JSON-RPC client → MCP server
→ tools/list request
→ Server returns tool schemas
→ Client calls search_context
→ Server returns results
→ Complete in < 500ms

Data Persistence Flow

• Create/update embeddings (10 testes)
• Cache locally (8 testes)
• Sync to cloud (8 testes)
• Retrieve from local cache (8 testes)
• Retrieve from cloud (8 testes)

SLA: Sync completes in < 5s for 100 docs

Data Persistence Scenario

User: Makes 5 searches offline
→ Results cached locally
→ Goes online
→ Data syncs to MongoDB
→ Other devices see updates
→ Sync completes in < 2s

Multi-step Complex Queries

• Analyze dependencies (8 testes)
• Find tests for code (8 testes)
• Generate documentation (8 testes)
• Suggest improvements (8 testes)

SLA: Complex query result in < 2s

Complex Query Scenario

User: "Show me all callers of getUserById and their tests"
→ Vectora analyzes dependencies
→ Finds all calling functions
→ Finds tests for those functions
→ Returns structured results
→ Complete in < 1.5s

E2E SLAs

Testing Approach

Tools

• Gemini CLI: Integration testing with mocked Vectora
• VS Code: UI testing with Playwright
• MCP: Protocol testing with JSON-RPC compliance
• Database: Transaction testing with MongoDB
• Performance: Load testing with 50+ concurrent users

Scenarios Covered

• Happy path (all systems working)
• Partial failures (one component down)
• Network issues (intermittent connectivity)
• Authentication failures (expired tokens)
• Rate limiting (quota exceeded)
• Data conflicts (concurrent modifications)

External Linking