Nova-7
Nova-7 — the interactive gateway to the knowledge base
Nova-7 is the assistant of agent-007.ai. She answers questions from the published content and guides you to the relevant missions and primary sources. This page openly explains how she works — and where her limits lie.
How Nova-7 works
Nova-7 is embedded in the site as a chat. Behind the scenes runs a workflow based on n8n that takes in your question, draws relevant context from this site's content, and has an answer generated from it. The actual language capability is currently provided by Google Gemini.
Grounded in the content Nova-7 answers from the bundled, published knowledge base of this site — all missions, architecture, radar, practice, glossary and sources. She points to the relevant page and says openly when something is not covered there, instead of guessing. The knowledge base is automatically updated with every publication.
Model choice The model used is deliberately interchangeable — that is exactly what Mission 006 (Portability) describes. As of today, Google Gemini generates the answers; switching the provider does not change how the site works.
Access — Nova-7 with bot protection
The chat is openly usable and protected from automated misuse of the model interface by a bot check (Cloudflare Turnstile). On your first message you briefly confirm you're not a bot — after that the chat stays unlocked for the session. Alternatively, an activation word still works (a short message to contact@agent-007.ai).
What Nova-7 is intended for
- Explaining the content of the seven missions, the reference architecture and the radar.
- Naming the pages and primary sources relevant to your question.
- Putting terms in context and establishing connections between the missions.
Example questions to get started
These questions show what Nova-7 is well suited for:
- What distinguishes a copilot from an agent?
- Why does an agent need its own identity?
- What role do evals play in productive operation?
- How are the seven missions connected?
- What does Human-in-the-Loop mean for agent systems?
- What architectural foundations does a production-ready AI agent need?
Limits — what Nova-7 is not
- No legal, tax or security advice. For regulation (e.g. the EU AI Act), the original text is authoritative.
- No infallible oracle. Answers can be wrong or incomplete — please check the cited sources.
- No access to your systems. Nova-7 only knows the public content of this site, not your environment.
- No live web. She does not research live, but works from the curated knowledge base.
Handling of sources
The content draws on recognised standards and primary sources (NIST, OWASP, IETF, ISO, OpenTelemetry, EU AI Act). Editorial assessments are marked as such. How sources are selected and maintained is described on the page Sources & Methodology.
Data protection in the chat
When you write something to Nova-7, your input is transmitted to the model provider (currently Google Gemini) to generate the answer. For this reason, please do not enter any personal, confidential or business-critical data in the chat.
How your inputs are handled Your chat inputs are not stored permanently in a database; for conversational context the workflow only keeps the last few messages transiently in memory. To generate the answer, the inputs are transmitted to Google Gemini as a processor.
Further information under Data Protection.