Settings & configuration
Top-level configuration for WP-PFAgent. Most settings live in the admin SPA under Setyenv → Agent → Settings.
Provider credentials
- API key — encrypted at rest (see Credential encryption).
- Provider presets — built-in catalog of the major providers and any OpenAI-compatible endpoint. See Provider presets.
- Per-model configuration — after the setup wizard, the model list, capability flags, and pricing for each model live alongside the credential and are visible in the model picker.
Rate limits
Per-user, per-provider sliding windows. Configurable in Settings → Rate limits:
| Bucket | Purpose |
|---|---|
| Reads | List/get calls into the agent surface. |
| Config changes | Provider, credential, and rules edits. |
| LLM calls | Total LLM round-trips. |
| Agent turns | Chat turns started by the user. |
Long conversations
When a conversation approaches the model's context limit, the agent keeps it flowing automatically — older history is summarised so recent turns stay intact. The behaviour is on by default; the trade-off and how to tune it live in the admin under Settings → Conversation length.
Provider-specific tuning
Some providers expose features that the agent can opt into per model — for example, response caching on providers that support it. These are surfaced as toggles next to the model in the picker so you can enable them where it helps and leave them off where it does not.