rag-search
AI search over the portals' documentation
People could not find answers in the documentation, and clients would not send their data to a third-party AI. So we built a search and a chat we host ourselves, which reply from each portal's own documentation and translate the content automatically.
- Qdrant and S3 Vectors behind a LiteLLM proxy
- SQS queues with DLQs and scheduled reconciliation
- Observability with Langfuse and OpenTelemetry