How does Timpi Orca support RAG workflows?
RAG workflows combine retrieval, context preparation and generation in a controlled automation flow, with storage options for fast cache, Redis, large OpenSearch-backed knowledge bases, or external RAG systems.
Answer
Timpi Orca supports RAG workflows by letting users connect retrieval, data, transform and AI model blocks. A workflow can retrieve background knowledge, merge it with a user question, call a suitable model and return an answer with the supporting context.
Orca supports four RAG configurations so teams can choose the right balance of speed, scale and ownership:
- In-memory cache: designed for blistering fast response times when the RAG data is small, hot and frequently reused.
- Redis: designed for very fast response times with a shared cache/vector layer that can support multiple controller instances.
- Orca-managed OpenSearch: designed for large RAG datasets where users want Timpi to provide and operate the storage environment.
- Bring your own RAG: connect an existing external RAG or vector database to the workflow when the customer already operates their own retrieval stack.
RAG can be combined with Timpi data sources, customer knowledge bases, external vector indexes and model routing. Orca also supports parallel branches, so retrieval can happen while another branch classifies intent or prepares model routing.
- RAG Chat blocks for retrieval plus generation
- Transform blocks for context merging and prompt shaping
- RAG Retrieve blocks for returning matching chunks and merged context
- Graph and report blocks for visualizing retrieved insights
- Published API endpoints for application integration