added README note on poppler install and better error handling for po…

…ppler not found

elm/pdf.py

@@ -254,12 +254,21 @@ def clean_poppler(self, layout=True):
             if not os.path.exists(os.path.dirname(fp_out)):
                 os.makedirs(os.path.dirname(fp_out), exist_ok=True)
 
-            stdout = subprocess.run(args, check=True, stdout=subprocess.PIPE)
-            if stdout.returncode != 0:
-                msg = ('Poppler raised return code {}: {}'
-                       .format(stdout.returncode, stdout))
+            try:
+                stdout = subprocess.run(args, check=True,
+                                        stdout=subprocess.PIPE)
+                if stdout.returncode != 0:
+                    msg = ('Poppler raised return code {}: {}'
+                           .format(stdout.returncode, stdout))
+                    logger.exception(msg)
+                    raise RuntimeError(msg)
+            except Exception as e:
+                msg = ('PDF cleaning with poppler failed! This usually '
+                       'because you have not installed the poppler utility '
+                       '(see https://poppler.freedesktop.org/). '
+                       f'Full error: {e}')
                 logger.exception(msg)
-                raise RuntimeError(msg)
+                raise RuntimeError(msg) from e
 
             with open(fp_out, 'r') as f:
                 clean_txt = f.read()

examples/energy_wizard/README.rst

@@ -8,6 +8,8 @@ corpus.
 
 Notes:
 
+- In this example, we use the optional `popper <https://poppler.freedesktop.org/>`_ PDF utility which you will have to install separately. You can also use the python-native ``PyPDF2`` package when calling using ``elm.pdf.PDFtoTXT`` but we have found that poppler works better.
+
 - Streamlit is required to run this app, which is not an explicit requirement of this repo (``pip install streamlit``)
 
 - You need to set up your own OpenAI or Azure-OpenAI API keys to run the scripts.