What's the expected source impedance? A meg or less?
What's the desired accuracy -- how much DC and noise error is tolerable?
What's the temperature range?
BAT54 is out, due to leakage even at room temperature, but PN diodes (BAV99, etc.) are open, at least as long as the required accuracy isn't stellar.
LEDs are indeed quite good, or they should be; the downside is, you're dependent on unspecified characteristics. (By the way, don't use blue LEDs! They are more ESD sensitive than red.)
Note that the leading 10k resistor is likely to arc over in case of ESD, so choose a diode or TVS capable of handling full surge current, whatever that may be.
If you need terribly high accuracy, consider using a buffer stage instead, and bootstrapping the ESD diodes. Then followed with the requisite gain stage.
Bootstrapped ESD diodes are wired as shown (typically with the antiparallel pair), but instead of a ground return, they connect to the buffer's output, through a small resistor. Supply clamp or TVS diodes then limit the buffered voltage (after the series resistor), so that incoming ESD goes through two diodes in series (the antiparallel pair, and the clamp), and not through the buffer's input or output structures (thanks to the series resistors).
Zeners or even GDTs may also be of interest, despite the low operating input voltage. A 12V zener (antiseries), placed where the LEDs are shown, will have quite low leakage at signal voltages (<6V say?), and the peak clamping voltage (say ~18V, plus ESL?) won't deliver too much current into the op-amp input pins, thanks to the second series resistor. (But mind that the injected current may well upset the amp's internal workings, leading to a long recovery tail, or hopefully not latchup, but you never know I guess.)
Tim