g = 2/3 # percolation: sqrt(q) = -2 cos(pi g) at q = 1
xT, xH = 3/(2*g) - 1, (2*g - 1)*(3 - 2*g)/(8*g)
nu = 1/(2 - xT)
[2 - 2*nu, nu*xH, nu*(2 - 2*xH), (2 - xH)/xH, nu, 2*xH] # alpha, beta, gamma, delta, nu, eta
[1/(nu*(2 - xH)), 1 + 2/(2 - xH), 2 - xH] # sigma, tau, d_fEvery entry is a rational computed from the Coulomb-gas coupling $g$ of its class, $\frac34$, $\frac56$, $1$ and $\frac23$ on the Potts branch and $\frac32$ on the $O(n)$ branch, through the scaling dimensions $x_T$ and $x_H$ and the relations in $d=2$, every division between Sage rationals; the scaling relations of Rushbrooke, Widom, Fisher and Josephson are asserted on every class before a value is returned, the percolation relations for $\sigma$, $\tau$ and $d_f$ on that class, and the Ising exponents are required to come out the same from $q=2$ and from $n=1$.
Exactness is not the same as proof: the Ising values and the percolation values of $\beta$, $\gamma$, $\nu$ and $\eta$ are theorems, and the 3-state and 4-state Potts values, the self-avoiding-walk values and the percolation $\alpha$ are exact conjectures of the Coulomb gas and conformal field theory, universally accepted and confirmed numerically. Before any entry was written, all 30 exponents $\alpha$ to $\eta$ were compared with the table on Wikipedia's page on universality classes and $\sigma$, $\tau$, $d_f$ with its page on percolation exponents; the thermal and magnetic dimensions of the Ising and 3-state Potts classes with the Kac formula of the minimal models, $h_{2,1}=\frac12$ and $h_{1,2}=\frac{1}{16}$ for $M(4,3)$ and $h_{2,1}=\frac25$ and $h_{3,3}=\frac{1}{15}$ for $M(6,5)$, and the thermal dimensions of percolation and of the walk with $h_{2,1}=\frac58$ and $h_{1,3}=\frac13$ at $c=0$, a computation sharing nothing with the Coulomb gas; the percolation values with the theorem of Smirnov and Werner and the one-arm exponent of Lawler, Schramm and Werner; $d_f=\frac{91}{48}$ with a Monte Carlo measurement of the largest cluster of site percolation on the triangular lattice at $p_c=\frac12$, which gave $1.88$ between $L=128$ and $L=2048$ with a control at $p=0.65$ returning $2.00$, and $\tau$ with the cluster-size distribution there, which gave $1.98$ to $2.01$ depending on the range fitted; and $\gamma=\frac{43}{32}$ and $\nu=\frac34$ of the self-avoiding walk with an exact enumeration of square-lattice walks to 16 steps, whose ratio estimates $1.340$ and $0.730$ rise towards the values, the ordinary random walk as a control returning exactly $1$ and $\frac12$.