Avaya System Evaluator — Admin Configuration
← Back to Tool
AI Guidance
AI Knowledge
Button Features
Scoring Thresholds
Platforms & Rules
H.323 Model Lookup
Users
Platforms
Select a platform to edit its AI knowledge.
Button Features Scanned in XLSX

These are the Avaya button feature codes the processor counts per extension. Match the VBA InitButtonFeatures list exactly.

⚠ Button feature changes require a service restart to take effect on new uploads.
Scoring Thresholds
Display Color Thresholds

Controls when counts on the summary page turn orange (elevated) or red (high). Set red to 0 to disable red highlighting.

NWN Partner Tier Bonuses

Points added to a platform's fit score based on its NWN partner tier. Set in Admin → AI Knowledge per platform.

Tier Descriptions
UC
CC
Both
Platforms
H.323 Model Lookup

Maps H.323 firmware release strings (from reg-stations_*.csv) to a display model name. Matching uses starts-with logic — e.g. a prefix of 6.8 matches 6.8.1, 6.8.0.1, and 6.8.9a. When multiple entries match, the longest prefix wins. Overridden names appear in red in the Excel report.

Release Prefix Model Override
Users

Accounts in the Cognito user pool backing this deployment's login. No self-service signup — new accounts are created via the AWS CLI (see infra/README.md "Managing users"). Last Login is sourced from CloudTrail's InitiateAuth history, which only retains ~90 days — a blank value means no sign-in in that window, not necessarily "never".

Email Status Admin Created Last Login
External AI Provider

External AI runs through Amazon Bedrock — IAM-authenticated, no API key to configure or rotate. Access is controlled entirely by the Lambda execution role's IAM policy; there's nothing to set here.

Internal AI Endpoint

Used when a user selects Internal AI on the upload form. Point this at any OpenAI-compatible chat endpoint hosted on-premises (Ollama, vLLM, LM Studio, etc.). Environment data never leaves the network when Internal AI is selected.

Must expose /chat/completions under this URL (include the /v1 suffix if the server uses one).

Analysis Mode — AI Evaluation Options

Check which choices appear in the AI Evaluation dropdown on the upload form (Step 2). These options match the eval page one-to-one. At least one must be checked. Changes take effect immediately — no restart required.

AI System Prompt

Configure the system prompt sent to Claude when analyzing each Avaya environment. These fields shape the AI's persona, communication style, assessment goal, step-by-step workflow, and guardrails. Active whenever the external AI provider (Amazon Bedrock) is reachable.

Who the AI is — role, expertise, and communication style foundation.

How the AI communicates — formality, directness, and emphasis guidelines.

What the AI should accomplish in each assessment — the primary objective.

Step-by-step instructions for how the AI should process and structure its analysis.

Hard constraints — what the AI must never do, reference, or speculate about.

nwn.ai Admin

Sign in with your account. You must be a member of the nwn-admins group.

Forgot password?