Rivine Chat

Rivine Chat is the general-purpose conversational engine, and it is deliberately the least specialized of the five. Its job is to run a full reasoning pass, sense the request, reason about what it actually needs, act (which may mean pulling in grounded data, reading a file, or delegating), and synthesize a final answer, inside a single continuous loop, without the multi-stage delegation machinery the more specialized engines use.

Figure: Chat shape: one loop, shared RE-AI and RCP underneath. Boxes name stages of work, not internal classes or tools.

What distinguishes Chat from "a chatbot with tools" is that it inherits the full RE-AI reasoning framework and the full RCP-governed data catalog: it has access to the same breadth of external data sources as the domain engines, and in live testing it demonstrably reaches for the right combination of them rather than defaulting to a single generic search path. A representative internal test, a question requiring both a specialized authoritative lookup and a general economic statistic, resulted in the engine chaining a multi-step lookup against the correct grounded source and falling back to general search only for the piece of the question that had no specialized source available. That is the self-heal and discovery layers of RCP working as intended, inside the simplest of the five engines.

┌──────────────┐     ┌─────────────────────────────┐     ┌──────────────┐
│ User request │────▶│  Chat continuous loop       │────▶│ Final answer │
└──────────────┘     │  sense → reason → act → out │     └──────────────┘
                     └──────────────┬──────────────┘
                                    │
                     ┌──────────────▼──────────────┐
                     │ RCP catalog  (specialized   │
                     │ sources first; search last) │
                     └─────────────────────────────┘