Settings / Self-Serve AI

Self-Serve AI Configuration

Controls the provider, model, and parameters behind Step 2 (business profile analysis), Step 4 (prompt suggestion generation), and the assessment run — for self-serve signups only.

No configuration saved yet — the fields below are pre-filled with system defaults (each adapter’s hardcoded constants). Nothing has been persisted; save a version once you’re ready to make these the source of truth.

These settings apply only to self-serve client signups (Client.source === "WEB"). Assessments you trigger manually for admin-created clients keep using each provider’s built-in defaults, unchanged. Grounding/web search stays permanently on for the assessment run and isn’t configurable here; repeatsPerPrompt stays fixed at 1.

Business Profile Analysis

lib/aiExtraction.ts

Step 2

Provider

Perplexity isn’t offered — structured-output reliability isn’t validated for it yet.

Model

Temperature

Not configurable for Claude — extended thinking requires default sampling.

Max tokens

Prompt Suggestion Generation

lib/promptGeneration.ts

Step 4

Provider

Perplexity isn’t offered — structured-output reliability isn’t validated for it yet.

Model

Temperature

0–1 for CLAUDE

Max tokens

Prompts per topic

Generated per fixed topic — Perbandingan, Kasus Penggunaan, Kategori Produk, Anggaran, Industri (5 topics). Total suggestions shown to the client = this number × 5.

Assessment Run Providers

lib/providers/* — per-provider override for the assessment run itself

Active	Provider	Model	Temperature	Max tokens
	OpenAI
	Claude
	Gemini
	Perplexity

At least one provider must be active.