Documentation & FAQ
Below you’ll find answers to the most common questions about Nova Rafale.
Click on a topic to expand the details.
General
1 · What is Nova Rafale in one sentence?
Nova Rafale is an EU‑hosted AI orchestration platform that unifies language, vision and audio models under a single secure workflow engine and API gateway.
2 · Which user profiles benefit most from Nova Rafale?
Marketing teams, developers, designers, data analysts, educators and indie creators rely on Nova Rafale to prototype content, code or design dramatically faster and cheaper than traditional methods.
3 · Who owns the generated output?
You do. All outputs come with full commercial rights for end‑users. Third‑party model terms (currently also unrestricted) still apply.
4 · Is there a desktop or mobile app?
Cross‑platform desktop (macOS/Windows/Linux) and a PWA mobile experience are available; both sync with the web console.
Account & Billing
5 · What payment methods are supported?
Visa, Mastercard, Amex, Sofort, SEPA, Apple Pay, Google Pay; yearly invoices payable by wire.
6 · Do prices include model usage costs?
Yes. Each subscription bundles a shared token pool across all integrated providers. High‑volume plans can add wholesale token blocks.
7 · Can I downgrade or upgrade anytime?
Plan changes take effect instantly with prorated credits or charges.
8 · How do I add teammates?
Invite colleagues via email and assign Viewer, Editor, Manager or Admin roles. Billing roles don’t count toward seat limits.
9 · Do you have student or non‑profit pricing?
Eligible organisations receive 35 % off all tiers—just submit proof of status.
AI Models & Quality
10 · Which language models are available?
GPT‑4o, Claude 3 Opus, Gemini Ultra plus Lumi‑7B (Nova Rafale’s distilled model) for low‑latency jobs.
11 · Can I lock requests to one provider?
Yes—select the desired model in the template’s Advanced tab and override temperature, top‑p and max‑tokens.
12 · How do you mitigate hallucinations?
Retrieval‑augmented generation, sentence‑level factuality scoring and automated citation insertion keep outputs grounded.
13 · Is image generation brand‑safe?
Yes. Dual‑stage NSFW and IP filters plus C2PA provenance metadata ensure compliance.
14 · Do you support fine‑tuning?
Private LoRA and OpenAI Fine‑Tuning v2 endpoints are in beta; bring your data or connect to our vector database.
Data & Security
15 · Where is data stored?
Hetzner (Frankfurt) primary with encrypted daily backups to Helsinki.
16 · How long are prompts stored?
Default 30 days (configurable to 0); logs are encrypted at rest.
17 · Which auth standards are supported?
OAuth 2.1, short‑lived JWTs, optional SAML SSO & SCIM for enterprise.
18 · Are DPAs and SCCs available?
Yes—download them anytime from Settings → Compliance.
19 · ISO 27001 status?
Audit underway; certificate expected Q4 2025. SOC2 Type II follows.
Integrations & API
20 · Is there an OpenAPI spec?
Grab it at
/docs/openapi.yaml—regenerated nightly.21 · Do you support streaming?
Yes—WebSockets or HTTP SSE for token‑level streaming.
22 · SDK availability?
Official Python, Node.js, Go, PHP; community Java, Ruby, .NET.
23 · Zapier/Make integration?
Pre‑built actions for Text, Image and Moderation—no code needed.
24 · What are the rate limits?
Free 20 req/min, Pro 120 req/min, Enterprise custom SLA.
Troubleshooting
25 · Job stuck in queue?
Check
status.nova-rafale.com, refresh token, then retry. It may indicate quota exhaustion or provider downtime.26 · Output truncated?
Increase max‑tokens or switch length mode to “long”.
27 · Garbled UTF‑8?
Enable UTF‑8 normalisation or set Mono‑Language to your locale.
28 · Content policy violation?
Remove restricted keywords like brand names, celebrities or NSFW terms.
29 · Voice‑over desync?
Lower speech rate or split paragraphs under 300 words.
Performance & Analytics
30 · How are costs calculated?
Provider token rates normalised to USD + 5 % platform fee.
31 · Export usage data?
CSV/JSON export or Stats API for BI tools.
32 · Cost caps available?
Set soft/hard limits with Slack or email alerts.
33 · Typical latency?
Text <40 k ≈ 720 ms, image 1024² ≈ 6 s; Enterprise GPU pin ≈ 450 ms.
34 · Model success stats?
Reliability tab shows success/fail & latency percentiles by provider.
Legal & Compliance
35 · Is content plagiarism‑free?
Built‑in originality checker flags similarity above 5 % against a 20 B page index.
36 · DMCA takedown process?
Designated agent responds within 24 h; offending content is quarantined.
37 · EU AI Act readiness?
Yes—transparency logs, opt‑outs and FRIA for high‑risk uses.
38 · Where is the DPA?
Workspace → Settings → Compliance. Signed PDF auto‑generated.
39 · Do you offer a Bug Bounty?
Yes—up to 5 000 USD per valid report via HackerOne.
40 · SLA terms?
Enterprise plans include a 99.9 % uptime SLA and priority support (<5 min median response).