\begin{table}[htbp]
\centering
\caption{Operational characteristics of the platform. Latencies are indicative order-of-magnitude
figures for the corresponding subsystems rather than benchmarked measurements; they motivate the
streaming and single-call design choices discussed in the text.}
\label{tab:robustness}
\begin{threeparttable}
\begin{tabular}{@{}lll@{}}
\toprule
Dimension & Mechanism & Indicative characteristic \\
\midrule
First-token latency   & SSE streaming                 & sub-second perceived response \\
Tool concurrency      & batched execution             & $\leq 5$ tools / batch, 500\,ms spacing \\
Context window        & adaptive summarization        & effectively unbounded history \\
Generation cap        & per-response limit            & 64{,}000 tokens \\
Python analysis       & sandboxed subprocess          & seconds per execution \\
PDF rendering         & headless browser              & several seconds \\
Slide compilation     & two-pass \LaTeX{}             & roughly ten seconds \\
Cost control          & local token estimation        & per-session accounting \\
Rate-limit safety     & inter-batch delay             & bounded provider load \\
\bottomrule
\end{tabular}
\begin{tablenotes}[flushleft]
\footnotesize
\item \textit{Notes.} Figures characterize the subsystems qualitatively to explain design
trade-offs; a controlled performance evaluation is left to future work
(Section~\ref{sec:conclusion}).
\end{tablenotes}
\end{threeparttable}
\end{table}