changed the structure

app.py

@@ -0,0 +1,91 @@
+import gradio as gr
+
+
+import src.control as ctrl
+
+
+"""
+==================================
+A. Component part
+==================================
+"""
+
+with gr.Blocks() as hrqa:
+
+    with gr.Row():
+
+        with gr.Column():
+            pass
+
+        with gr.Column(scale=10):
+            """
+            1. input docs components
+            """
+
+            gr.Markdown("# Questions sur le vivre ensemble en entreprise")
+
+            input_text_comp = gr.Textbox(
+                label="",
+                lines=1,
+                max_lines=3,
+                interactive=True,
+                placeholder="Posez votre question ici",
+            )
+            input_example_comp = gr.Radio(
+                label="Examples de questions",
+                choices=["Remboursement de frais de voiture", "Recommandations de transport"],
+            )
+            output_text_comp = gr.Textbox(
+                label="La réponse automatique",
+                lines=2,
+                max_lines=10,
+                interactive=False,
+                visible=False,
+            )
+            sources_comp = gr.CheckboxGroup(
+                label="Documents sources",
+                visible=False,
+                interactive=False,
+            )
+
+        with gr.Column():
+            pass
+
+
+    def input_text_fn1():
+        update_ = {
+            output_text_comp: gr.update(visible=True),
+        }
+        return update_
+
+    def input_text_fn2(input_text_):
+        answer, sources = ctrl.get_response(query=input_text_)
+        source_labels = [s['distance']+' '+s['paragraph']+' '+s['title']+' from '+s['doc'] for s in sources]
+        update_ = {
+            output_text_comp: gr.update(value=answer),
+            sources_comp: gr.update(visible=True, choices=source_labels, value=source_labels)
+        }
+        return update_
+
+
+    def input_example_fn(input_example_):
+        examples = {
+            "Remboursement de frais de voiture": "Comment sont remboursés mes frais kilométriques sur mes trajets "
+                                                 "professionnels?",
+            "Recommandations de transport": "Quelles sont les recommandations de l'entreprise? Vaut-il mieux voyager en "
+                                            "train ou en avion?"
+        }
+        update_ = {
+            input_text_comp: gr.update(value=examples[input_example_]),
+            output_text_comp: gr.update(visible=True),
+        }
+        return update_
+
+    input_text_comp\
+        .submit(input_text_fn1, inputs=[], outputs=[output_text_comp])\
+        .then(input_text_fn2, inputs=[input_text_comp], outputs=[output_text_comp, sources_comp])
+    input_example_comp\
+        .change(input_example_fn, inputs=[input_example_comp], outputs=[input_text_comp, output_text_comp])\
+        .then(input_text_fn2, inputs=[input_text_comp], outputs=[output_text_comp, sources_comp])
+
+hrqa.queue().launch()

src/control/control.py

@@ -0,0 +1,49 @@
+import chromadb
+
+import src.tools.retriever as rtrvr
+import src.tools.llm as llm
+from src.domain.doc import Doc
+
+chroma_client = chromadb.Client()
+
+plan_language = 'en'
+content_language = 'en'
+path_plan = '../data/business_trips_plan_until_9_en.docx'
+path_content = '../data/business_trips_content_until_9_en.docx'
+collection_name = "until_9"
+
+doc_plan = Doc(path_plan)
+doc_content = Doc(path_content)
+collection_ = rtrvr.init_collections(chroma_client, doc_plan, doc_content, collection_name)
+
+
+def get_response(query):
+    if plan_language == 'en':
+        query = llm.translate(query)
+    sources = rtrvr.similarity_search(collection=collection_, query=query)
+    sources = select_best_sources(sources)
+    sources_contents = [s['content'] for s in sources]
+    context = '\n'.join(sources_contents)
+    answer = llm.generate_paragraph(query=query, context=context, language=content_language)
+    if content_language == 'en':
+        answer = llm.translate(text=answer, language='fr')
+    return answer.lstrip(), sources
+
+
+def select_best_sources(sources: [], delta_1_2=0.1, delta_1_n=0.25, absolute=1.1) -> []:
+    best_sources = []
+    for idx, s in enumerate(sources):
+        if idx == 0 \
+           or (s['distance_f'] - sources[idx - 1]['distance_f'] < delta_1_2
+                and s['distance_f'] - sources[0]['distance_f'] < delta_1_n) \
+           or s['distance_f'] < absolute:
+            best_sources.append(s)
+    return best_sources
+
+
+q1 = "Comment sont remboursés mes frais kilométriques sur mes déplacements avec mon véhicule personnel?"
+q2 = "Quels sont les moyens de transport recommandés par la société?"
+q3 = "est-ce que mes billets de cinéma peuvent être remboursés?"
+
+a2 = get_response(q3)
+print(a2)