B-model and Hochschild cohomology

The boundary conditions of the B-model, ie, the D-branes, are the objects in $\mathcal{D}^b(X)$. A little bit of playing with pictures gives that the space of closed string states must be in the center of the algebra of open strings for any given boundary condition. For a category $\mathcal{C}$, this gives a map to $\mathcal{Nat}(id_\mathcal{C},id_\mathcal{C})$. Extending this to the graded situation and using the description of natural transformations as bimodules, you get $$ HH^\bullet(\mathcal{D}^b(X)) = \mathrm{Ext}^\bullet_{X \times X}(\mathcal{O}_\Delta,\mathcal{O}_\Delta) $$ where $\Delta$ is the diagonal. There's a version of the HKR theorem that relates this to cohomology of exterior powers of the tangent bundle, the usual space of states in the B-model.

A lot of this is explained in Moore and Segal. I also explained some (with no claim to originality) in the beginning of my old paper Deformations and D-branes.

For correct attribution, one should at least mention the paper which the preprint of Moore and Segal itself quotes as the source for the particular case of the algebraic description of open-closed TFTs which they give without proof. The following paper was published about 5 years before the preprint of Moore and Segal:

C. I. Lazaroiu, On the structure of open-closed topological field theory in two dimensions, Nucl.Phys.B603 (2001) 497-530, https://arxiv.org/abs/hep-th/0010269

Notice that the spaces of boundary states in the B-model with CY manifold target X are Z-graded, since the physically correct category of B-type topological D-branes is not D^b(X) but rather its shift completion. The Z/2Z grading in Lazaroiu's axioms is the mod 2 reduction of that natural Z-grading -- and this Z/2Z grading is ignored by Moore and Segal. Shift completions were explained in:

C. I. Lazaroiu, Graded D-branes and skew-categories, JHEP 0708 (2007) 088, https://arxiv.org/abs/hep-th/0612041

(which also explains the twist-completion that occurs, for example, in B-type orbifold LG models).

One should also note that the argument using HH(D^b(X)) and the holomorphic HKR isomorphism is not due to Lurie but to Kontsevich. Furthermore this argument (while very nice and natural) is not needed for the B-model, since the closed string state space of the B-model is known independently from localization of the path integral in the bulk sector, which produces the space of polyvector fields endowed with its (Dolbeault) differential and Serre trace. This goes back to Witten and to the paper of BCOV in the early 1990's. The advantage of the polyvector approach is that it provides an off-shell model which also appears in the BCOV theory and has a dual description using a BV bracket.