1: \begin{abstract}

2: 

3: The geometric Langlands program is distinguished in assigning

4: spectral decompositions to all representations,

5: not only the irreducible ones.

6: However, it is not even clear what is meant by a spectral decomposition

7: when one works with non-abelian reductive groups and with 

8: ramification.

9: The present work is meant to clarify the situation, showing

10: that the two obvious candidates for such a notion 

11: coincide. 

12: 

13: More broadly, we study the moduli space of (possibly irregular) de Rham

14: local systems from the perspective of homological algebra.

15: We show that, in spite of its infinite-dimensional nature, this

16: moduli space shares many of the nice features of

17: an Artin stack. Less broadly, these results support the belief in

18: the existence of a version of geometric Langlands allowing

19: arbitrary ramification.

20: 

21: Along the way, we give some apparently new, if unsurprising,

22: results about the algebraic geometry of 

23: the moduli space of connections, using Babbitt-Varadarajan's

24: reduction theory for differential equations.

25: 

26: \end{abstract}

27: