Fixed a bunch of stuff

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import paperqa
 import pickle
+import pandas as pd
 from pathlib import Path
 import requests
 import zipfile
@@ -9,27 +10,35 @@ import tempfile
 import os
 
 
-docs = None
+css_style = """
 
+.gradio-container {
+    font-family: "IBM Plex Mono";
+}
+"""
 
-def request_pathname(files):
+
+def request_pathname(files, data, openai_api_key):
     if files is None:
         return [[]]
-    return [[file.name, file.name.split('/')[-1], None] for file in files],  [[len(files), 0]]
+    for file in files:
+        # make sure we're not duplicating things in the dataset
+        if file.name in [x[0] for x in data]:
+            continue
+        data.append([file.name, None, None])
+    return [[len(data), 0]], data, data, validate_dataset(pd.DataFrame(data), openai_api_key)
 
 
 def validate_dataset(dataset, openapi):
-    global docs
-    docs = None  # clear it out if dataset is modified
     docs_ready = dataset.iloc[-1, 0] != ""
     if docs_ready and type(openapi) is str and len(openapi) > 0:
         return "✨Ready✨"
     elif docs_ready:
-        return "⚠️Waiting for key..."
+        return "⚠️Waiting for key⚠️"
     elif type(openapi) is str and len(openapi) > 0:
-        return "⚠️Waiting for documents..."
+        return "⚠️Waiting for documents⚠️"
     else:
-        return "⚠️Waiting for documents and key..."
+        return "⚠️Waiting for documents and key⚠️"
 
 
 def make_stats(docs):
@@ -41,18 +50,17 @@ def do_ask(question, button, openapi, dataset, length, do_marg, k, max_sources,
     passages = ""
     docs_ready = dataset.iloc[-1, 0] != ""
     if button == "✨Ready✨" and type(openapi) is str and len(openapi) > 0 and docs_ready:
-        if docs is None:  # don't want to rebuild index if it's already built
-            import os
-            os.environ['OPENAI_API_KEY'] = openapi.strip()
+        os.environ['OPENAI_API_KEY'] = openapi.strip()
+        if docs is None:
             docs = paperqa.Docs()
-            # dataset is pandas dataframe
-            for _, row in dataset.iterrows():
-                try:
-                    docs.add(row['filepath'], row['citation string'],
-                             key=row['key'], disable_check=True)
-                    yield "", "", "", docs, make_stats(docs)
-                except Exception as e:
-                    pass
+        # dataset is pandas dataframe
+        for _, row in dataset.iterrows():
+            try:
+                docs.add(row['filepath'], row['citation string'],
+                         key=row['key'], disable_check=True)
+                yield "", "", "", docs, make_stats(docs)
+            except Exception as e:
+                pass
     else:
         yield "", "", "", docs, [[0, 0]]
     #progress(0, "Building Index...")
@@ -71,10 +79,9 @@ def do_ask(question, button, openapi, dataset, length, do_marg, k, max_sources,
     yield result.formatted_answer, result.context, passages, docs,  make_stats(docs)
 
 
-def download_repo(gh_repo, pbar=gr.Progress()):
+def download_repo(gh_repo, data, openai_api_key, pbar=gr.Progress()):
     # download zipped version of repo
     r = requests.get(f'https://api.github.com/repos/{gh_repo}/zipball')
-    files = []
     if r.status_code == 200:
         pbar(1, 'Downloaded')
 
@@ -101,18 +108,22 @@ def download_repo(gh_repo, pbar=gr.Progress()):
                     rel_path = '/'.join(f.split('/')[1:])
                     key = os.path.basename(f)
                     citation = f'[{rel_path}](https://github.com/{gh_repo}/tree/main/{rel_path})'
-                    files.append([path, citation, key])
-                    yield files, [[len(files), 0]]
+                    if path in [x[0] for x in data]:
+                        continue
+                    data.append([path, citation, key])
+                    yield [[len(data), 0]], data, data, validate_dataset(pd.DataFrame(data), openai_api_key)
                 pbar(int((i+1)/len(z.namelist()) * 99),
                      f'Added {f}')
         pbar(100, 'Done')
     else:
         raise ValueError('Unknown Github Repo')
+    return data
 
 
-with gr.Blocks() as demo:
+with gr.Blocks(css=css_style) as demo:
 
     docs = gr.State(None)
+    data = gr.State([])
     openai_api_key = gr.State('')
 
     gr.Markdown(f"""
@@ -147,7 +158,7 @@ with gr.Blocks() as demo:
             headers=["filepath", "citation string", "key"],
             datatype=["str", "str", "str"],
             col_count=(3, "fixed"),
-            interactive=True,
+            interactive=False,
             label="Documents and Citations",
             overflow_row_behaviour='paginate',
             max_rows=5
@@ -165,9 +176,9 @@ with gr.Blocks() as demo:
     dataset.change(validate_dataset, inputs=[
                    dataset, openai_api_key], outputs=[buildb])
     uploaded_files.change(request_pathname, inputs=[
-                          uploaded_files], outputs=[dataset, stats])
+                          uploaded_files, data, openai_api_key], outputs=[stats, data, dataset, buildb])
     download.click(fn=download_repo, inputs=[
-                   gh_repo], outputs=[dataset, stats])
+                   gh_repo, data, openai_api_key], outputs=[stats, data, dataset, buildb])
     query = gr.Textbox(
         placeholder="Enter your question here...", label="Question")
     with gr.Row():

requirements.txt

@@ -1,4 +1,4 @@
-paper-qa>=0.0.18
+paper-qa>=0.0.20
 gradio
 requests
 transformers