Ecoar AI: Qualitative Research & Interview Platform with Ent ORM & pgvector
Qualitative research platform built in Go for conducting automated user interviews and thematic transcript analysis. Features multi-tenant i...
Organization / Context: Projeto Autônomo / Pesquisa Aplicada
Timeline: 2025 - Present
Domain Category: Data Science
Technologies: Go, Ent ORM, PostgreSQL, pgvector, Docker, TypeScript, React, Tailwind CSS, OpenAI API, Gemini API
Executive Overview
Qualitative research platform built in Go for conducting automated user interviews and thematic transcript analysis. Features multi-tenant isolation via Ent ORM and PostgreSQL RLS, semantic vector search using pgvector, flexible LLM provider management (BYOK), and asynchronous session transcription pipelines.
Architecture & Data Flow
The diagram below illustrates the sanitized high-level component topology and data processing pipeline:
flowchart TD
A[Respondente / Pesquisador] --> B[Interface Web & Agente Interativo]
B --> C[Backend Go / Ent ORM Privacy Layer]
C --> D[Banco Relacional PostgreSQL + Row-Level Security]
C --> E[Módulo de Embeddings & Busca Vetorial pgvector]
C --> F[Conectores LLM / Provedores BYOK]
F --> G[Agrupamento Temático & Síntese Qualitativa]
G --> H[Relatórios Analíticos & Transcrições Indexadas]
Key Engineering Highlights
- Three-Layer Tenant Isolation: Multi-tenancy enforced across HTTP middleware, Ent Privacy policies, and PostgreSQL Row-Level Security.
- Semantic Search with pgvector: Vector embeddings indexing interview transcript excerpts for semantic retrieval and thematic clustering.
- BYOK Architecture: Agency-level API key management with automated fallback and pluggable LLM provider presets.