The question of wavelength adjustment for fluorescence is a recurrent one and one day I'll have to make something because it's a real issue people are facing.
To comment on what has been said, yes indeed larger excitation wavelengths will shift the signal in the NIR part of the spectra where CMOS sensors are typically less sensitive. Also, the Raman effect itself decreases as the 4th power of the wavelength so going from 532 to 785 means you have about 5x less signal in the first place. Added to the ultra poor quantum efficiency of CMOS sensor above 900 nm, you end up with a system that is usually limited to 2000 cm-1 and which has about 100x less electrons generated compared to 532 nm. On the other hand, at 785 nm, fluorescence is gone. 630-650 nm excitation is a tradeoff in that search for low fluorescence because most fluorescence is already gone and you're still in a relatively good range of QE of the CMOS sensor (check the specific QE plot for the IMX265 sensor monochrome). Also, the Raman effect at 630 is only 2x smaller than at 532 nm so it's not as bad as 785 nm.
630-650 nm isn't such a bad choice and, to be honest, if I had to start the OpenRAMAN project now it's not sure that I would pick 532 nm again. Performant lasers are cheaper in that range, stock filters are also more performant (they won't cut at 450 cm-1 but below that) and because you spread more in the nm range you also make the spectrometer part a touch easier in terms of resolution. A few years ago, my boss even asked me to make a 635 nm version of OpenRAMAN for the company (before I was kicked out of my own project...). So I know it works.
What I can recommend is:
check all lasers with FWHM < 0.2 nm you can find in the red region (630-650 nm)
check where is the actual cutoff using stock dichroic mirrors (thorlabs or other) using their provided transmission/reflection chart
pick the laser and dichroic mirror that gives the best results
adjust the spectrometer angle to accomodate for the wavelength change.
check that the camera sensor is still the good one; at the office I had to upgrade to the 23S6M sensor but an alternative is to try using a 35 mm objective instead of a 50 mm one
Generally speaking, switching to lens with optical coating "B" will improve light transmission (instead of "A" currently used) and check the grating of edmund optics around ~1000 lp/mm (don't forget to take into account the blaze angle).
The tricky part is adjusting the spectrometer angles; not because of the maths but because it requires making a new baseplate (--> new optical design, reimport in solidworks, re-mate all the surfaces etc.) which is a lot of work. Technically, it's a 2-3 days job in total (so not that much) but I can't afford it at the moment... So it has been on my high priority list until I can find budget & time to make the work for a while now.
On your side, you can use breadboard components to experiment, or, more easy, you can machine new holes directly into the existing baseplate to relocate the camera to the proper angle. The cover won't work but a dark piece of clothes will also do the job!
I hope that gives you some insight on how to do the adaptation and if it worth it (spoiler alert: yes it does ).
My suggestion (take with a grain of salt):
Here's what you would need:
A new laser (here's options from my supplier): https://www.civillaser.com/index.php?ma ... &cPath=705
3-5ish new lenses (the lenses have coating for 532nm so check if new ones are needed besides the "main ones": dichroic, fiber filter, longpass filter)
You would move the lens group around so the large doublet is now focusing the light to a fiber. You'd have to adjust for focus. You could use the spec to find the spot that maximizes the light received but it should be close to the effective focal length of the doublet. The small doublet would be unused.
A S1SMA plate that would receive the light
A new ccd spectrometer. You can hunt around on alibaba express for a used or cheap old one and make sure it's sensitive in the new output range.