{"id":"openai-chatgpt-retrieval-plugin","name":"chatgpt-retrieval-plugin","af_score":37.2,"security_score":58.0,"reliability_score":21.2,"what_it_does":"A self-hosted Retrieval-Augmented Generation (RAG) backend that exposes a FastAPI service for semantic search over user-provided documents. It chunks documents, creates embeddings via OpenAI, stores/query them in a chosen vector database provider, and serves query/upsert/delete endpoints intended to be used as a ChatGPT Retrieval Plugin backend (and for custom GPT actions/function calling).","best_when":"You want a self-hosted, configurable RAG retrieval layer with pluggable vector database backends and metadata filtering, and you can manage hosting/operations and API authentication for your environment.","avoid_when":"You cannot guarantee secure deployment of the FastAPI server (auth, network controls) or you need strong multi-tenant/least-privilege authorization semantics beyond the provided token-based approaches.","last_evaluated":"2026-03-29T13:09:04.899977+00:00","has_mcp":false,"has_api":true,"auth_methods":["None","API key (Basic)","API key (Bearer)","OAuth (per README)","Bearer token via BEARER_TOKEN environment variable (for local/server setup)"],"has_free_tier":false,"known_gotchas":["Shared-token auth can complicate agent workflows in multi-tenant contexts if you need per-user separation.","The service relies on external vector database configuration and correct environment variables; misconfiguration may cause failures that agents can’t automatically remediate.","No explicit guidance is provided (in the excerpt) about pagination, rate-limit response headers, or safe retry/idempotency semantics for upsert/query operations."],"error_quality":0.0}