A good place to start might be Eppy. There's a good high-level introduction to what Eppy is and can do on Github.

Also, something the author, Santosh Philip said in a release email (April 2014) about Eppy's background:

"Eppy was developed over the last year, for the ASHRAE 1651-RP research project that required modifying large number of files and analyzing the results.

Eppy was partially funded by ASHRAE and the original code was written in 2004 as a text interface for Energyplus and the scripting language was built on it's foundations."

The open source license was GPLv3, but as of v0.5.0 it is MIT.

OpenStudio also has Python bindings, and works at a higher level of abstraction.

I use the OpenStudio Python bindings. You can build the initial model using the OpenStudio model and then convert to an IDF model with the built in translators. Any simulation components that are not directly available within OpenStudio can be added within the IDF model. The interface allows model objects to be queried using simple Python iterators.

zones = [zone for zone in openstudio.model.getThermalZones(model) if "Office" in zone.name().get()]

This makes manipulating models and running parametrics fairly simple. Also, the development team is continuing to improve the tool at a fast pace. There is excellent support for the product through Unmet Hours.

The downside is that you need to compile the bindings yourself. In Windows this requires working with Visual Studio. There is some guidance to compiling the bindings here. Getting the various paths set up properly is not always straight forward. See Python SDK Bindings for OpenStudio.

Here's a function I use for editing an IDF or IMF file in place that might be useful. It's inspired by the approach in jEPlus where you mark the variables you want to change in your IDF or IMF file by giving them a name in the form @@VariableName@@. It's handy for handling parametric runs where you are generating the values in Python code. Generally though it's easier to call jEPlus directly from Python (or not using Python at all...) if you just want a pre-set list of values or values from a known distribution.

import fileinput

def edit_template(substitutions, imf_path):
    """
    Replace all placeholder @@variables@@ in a template file with the
    correct values for the current run.

    Parameters
    ----------
    substitutions: dict
        A dictionary mapping the variable to be substituted to the value
        to be substitute in its place, e.g. {"BoilerEfficiency": 0.9} 
        will replace @@BoilerEfficiency@@ with 0.9 in the input file.
    imf_path : str
        The path to the IDF or IMF.

    Raises
    ------
    ValueError
        Raised if a value to replace was not found in the template file.

    """
    for key in substitutions:
        if key not in open(imf_path, 'rb').read():
            raise ValueError("%s not in %s" % (key, imf_path))
        for line in fileinput.FileInput(imf_path, inplace=True):
            for s in substitutions:
                 line = line.replace("@@%s@@" % s, str(substitutions[s]))  # enforce str
            print line,

Eppy is a really useful tool with a lot of thought gone into it as well so I wouldn't go reinventing the wheel. I'm sure any contributions you'd like to make would be welcomed. It's quite large, but the introductory tutorial is helpful (although if you already speak Python it can be a bit too detailed).

I haven't used the OpenStudio Python bindings so can't comment on that.

pyidf also became available several weeks ago. Haven't tried it and don't know how it differs from eppy, but it's there.