This article introduces the Electric Vehicle Itinerary Planning with Activities (EV-IPA) problem, which jointly optimizes vehicle routing, parallel activity scheduling, non-linear charging decisions, and time-window constraints. This tight coupling is increasingly relevant as conversational in-car assistants let drivers negotiate itineraries in natural language. However, directly embedding continuous, non-linear state-of-charge dynamics into a discrete decision space yields non-convex models incompatible with the sub-ten-second latencies demanded by interactive systems. We propose a Constraint Programming (CP) framework that decouples offline physics evaluation from online combinatorial search. Its core is a Dynamic Linear Energy Approximation that abstracts battery consumption into segment-wise bounds. Enforced via a structural Segment-Length Condition, this natively guarantees state-of-charge feasibility at every continuous point along the route. Furthermore, a hierarchical colocation graph captures parallel activities (e.g., charging while dining), and a precomputed convex envelope linearizes non-linear, temperature-dependent charging curves. On EV-IPA-1000, a new open benchmark of 1,000 European trip-planning instances, the framework produces itineraries with a median end-to-end latency of 1.58 seconds on production-grade data. By actively synchronizing charging events with user activities, the colocation architecture reduces average trip times by 7.9% on activity-bearing instances. Validation against a production-grade powertrain model confirms the fidelity of the linear abstractions: a 100% energy feasibility rate with mean state-of-charge and charge-gain deviations of 1.60 and 0.81 percentage points. The same rate is obtained on the open benchmark under its published surrogate model, and 99.9% is retained under a stress configuration that removes all soft safety margins.