Semantic search
Retrieve content by meaning, which helps when a query and its best source use different language.
Enterprise AI search
Find the document even when you do not remember its exact words
People rarely remember the exact title of a document or the wording inside it. Pūnaha searches by meaning to help them reach relevant passages for the question in front of them.
View original explanatory diagram
A better search experience
Help people get past “I know it is here somewhere”
Useful search has to cope with different wording, exact identifiers and knowledge spread across more than one kind of file.
Relevant passages
Return indexed passages ranked by vector similarity to the question, so you can inspect the material used as context.
Document understanding
Extract searchable text from supported office documents, PDF, email, RTF, EPUB and OpenDocument files.
Tenant-aware knowledge
Keep separate knowledge databases for different tenants and operational contexts.
Explore access governanceAI-ready context
Pass relevant retrieved material into a workflow so the selected model works with approved context.
Explore workflowsInspectable operation
Use application logs, audit information and workflow execution history to understand what happened.
Behind the result
What happens between a source file and an answer
Bring in supported content
Ingest text and supported files, including email and Outlook attachments.
Prepare the knowledge
Extract and organise the material for the selected tenant-aware knowledge store.
Retrieve relevant passages
Use semantic/vector retrieval to find passages relevant to the question.
Use the result
Return relevant context, support an AI response or continue through a governed workflow.
After the search
Finding the source is often only the first step
The information may need to support a response, a decision or a longer piece of work.
A retrieval step can sit inside a versioned workflow alongside prompts, model calls, approvals and merged results.
This lets teams reuse the same governed knowledge in different processes without forcing every task into a single assistant experience.
Direct knowledge discovery
Help a person locate material relevant to a question or investigation.
Retrieval-assisted AI
Provide a model with selected organisational context before it prepares a response.
Workflow input
Use retrieved information as one step in a longer process with review and action.
Questions
Enterprise AI search questions
What is the difference between semantic and keyword search?
Keyword search looks for matching terms. Semantic search looks for related meaning. Pūnaha currently uses vector-based semantic retrieval to find relevant passages.
Can search use private organisational content?
Yes. Pūnaha works with organisational knowledge and can run on customer-controlled infrastructure. Access design and source permissions still need to be configured for the intended environment.
Does Pūnaha support image-only PDF documents?
Text extraction is supported for PDFs that contain readable text. OCR for image-only PDFs is not currently presented as a ready website capability.
Can retrieved knowledge be used in an AI workflow?
Yes. Retrieval is a workflow building block and can provide relevant context to later prompt, model, approval or merge steps.
Product documentation
Try the knowledge and search steps
Use the knowledge and search user guide to see how to work with knowledge in the product. Keep source permissions and tenant access part of your evaluation.
Bring a question your current search struggles with
We can use a private demonstration to look at the source formats, search approach and next step you need to support.