Source to retrieval
Confirm extraction, tenant separation, relevance and the treatment of source permissions.
For technical buyers
Assess the architecture, integrations and control boundaries clearly
Review where Pūnaha runs, how it retrieves knowledge, which AI providers it supports, how workflows change and where a person can remain in the process.
Architecture
Controls
Operational fit
Capability view
What is ready to evaluate
| Area | Current capability | Evaluation focus |
|---|---|---|
| Deployment | Customer-controlled infrastructure or Pūnaha Cloud; Windows and Linux | Network, identity, operations and data path |
| Knowledge | Tenant-aware stores with semantic and hybrid retrieval | Sources, permissions, relevance and currency |
| Models | Local GGUF plus supported OpenAI, compatible, Azure OpenAI, Anthropic and Gemini connections | Provider terms, data path, performance and fit |
| Workflows | Validation, versioning, publication, parallel stages, approval and durable execution | Change control, failure handling and accountability |
| Extensions | Generic REST, read-only MySQL and generic web search backend | Authentication, allow-listing and least privilege |
| Operations | Application logs, audits and workflow execution history | Monitoring, retention and support model |
Evaluation areas
Test the whole path, not just the answer
Retrieval to model
Understand which content reaches which configured deployment under each workflow path.
Model to decision
Evaluate validation, approval, rejection and the handling of incomplete or unsuitable results.
Decision to operation
Review versioning, execution history, support boundaries and operational ownership.
Current boundaries
Know what is not being claimed
Clear limits make technical evaluation more useful.
This site does not present native Amazon Bedrock requests, complete OIDC sign-in, local MFA, OCR for image-only PDFs, active object-storage delivery or cloud-monitoring delivery as ready capabilities.
It also avoids a broad air-gap claim. Customer-managed deployment and local model support are available, but the complete environment must be assessed against the intended architecture.
Identity
Confirm the implemented sign-in approach and required enterprise identity integration for the target release.
Observability
Plan how application logs and workflow history fit the customer's monitoring approach.
Storage and migration
Confirm supported stores, delivery paths and any migration requirement before implementation.
Suggested evaluation
A focused technical review
Describe the environment
Document infrastructure, networks, data sources, model services and operational owners.
Choose a representative workflow
Include retrieval, a model decision and human approval if those elements matter to the intended use.
Test controls and failure paths
Review what happens when a source, model, approval or external service is unavailable or unsuitable.
Agree the implementation boundary
Record what Pūnaha provides and what remains part of the customer's surrounding platform and governance.
Bring your architecture questions to a technical demonstration
We can focus on deployment, retrieval, model connections, workflows and the current product boundaries.