Spaces:
Sleeping
Sleeping
File size: 16,949 Bytes
56a3465 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 |
import streamlit as st
from textwrap import dedent
from utils.audit.rag import get_text_from_content_for_doc,get_text_from_content_for_audio
from utils.audit.response_llm import generate_response_via_langchain
from langchain_core.messages import AIMessage, HumanMessage
from st_copy_to_clipboard import st_copy_to_clipboard
from utils.kg.construct_kg import get_graph
from audit_page.knowledge_graph import *
import json
import clipboard
from time import sleep
def graph_doc_to_json(graph):
nodes = []
edges = []
for node in graph.nodes:
node_id = node.id.replace(" ", "_")
label = node.id
type = node.type
nodes.append({"id": node_id, "label": label, "type": type})
for relationship in graph.relationships:
source = relationship.source
source_id = source.id.replace(" ", "_")
target = relationship.target
target_id = target.id.replace(" ", "_")
label = relationship.type
edges.append({"source": source_id, "label": label, "cible": target_id})
return {"noeuds": nodes, "relations": edges}
def chat_history_formatter(chat_history):
formatted_chat = ""
for message in chat_history:
if isinstance(message, AIMessage):
formatted_chat += f"AI:{message.content}\n\n"
elif isinstance(message, HumanMessage):
formatted_chat += f"Human:{message.content}\n\n"
return formatted_chat
def filter_correspondance(source_list:list[str],ref_dict:dict,reverse=False):
source_list = [item.lower().strip() for item in source_list]
if reverse:
return [key for key, value in ref_dict.items() if value.lower().strip() in source_list]
else:
# st.write(source_list)
# st.write(ref_dict.keys())
return [value for key, value in ref_dict.items() if key.lower().strip() in source_list]
@st.fragment()
def radio_choice():
options = ["compte_rendu","graphe de connaissance"]
choice = st.radio("Choisissez une option",options,index=st.session_state.radio_choice,horizontal=True,label_visibility="collapsed")
sleep(1)
if choice and options.index(choice) != st.session_state.radio_choice:
sleep(1)
st.session_state.radio_choice = options.index(choice)
return choice
def doc_dialog_main():
st.title("Dialogue avec le document")
if "audit" not in st.session_state or st.session_state.audit == {}:
st.error("Veuillez d'abord effectuer un audit pour générer le compte rendu ou le graphe de connaissance.")
return
#init cr and chat history cr
if "cr" not in st.session_state:
st.session_state.cr = ""
if "cr_chat_history" not in st.session_state:
st.session_state.cr_chat_history = [
]
#init graph and filter views
if "graph" not in st.session_state:
st.session_state.graph = None
if "filter_views" not in st.session_state:
st.session_state.filter_views = {}
if "current_view" not in st.session_state:
st.session_state.current_view = None
if "node_types" not in st.session_state:
st.session_state.node_types = None
# if "summary" not in st.session_state:
# st.session_state.summary = None
if "chat_graph_history" not in st.session_state:
st.session_state.chat_graph_history = []
#init a radio button for the choice
if "radio_choice" not in st.session_state:
st.session_state.radio_choice = None
# if "choice" not in st.session_state:
# st.session_state.choice = st.radio("Choisissez une option",["compte_rendu","graphe de connaissance"],index=st.session_state.radio_choice,horizontal=True,label_visibility="collapsed")
# choice = radio_choice()
options = ["compte_rendu","graphe de connaissance"]
choice = st.radio("Choisissez une option",options,index=st.session_state.radio_choice,horizontal=True,label_visibility="collapsed")
if choice and options.index(choice) != st.session_state.radio_choice:
st.session_state.radio_choice = options.index(choice)
audit = st.session_state.audit_simplified
content = st.session_state.audit["content"]
if audit["type de fichier"] == "pdf":
text = get_text_from_content_for_doc(content)
elif audit["type de fichier"] == "audio":
text = get_text_from_content_for_audio(content)
elif audit["type de fichier"] == "text":
text = content
prompt_cr = dedent(f'''
À partir du document ci-dessous, générez un compte rendu détaillé contenant les sections suivantes :
2. **Résumé** : Fournissez une synthèse complète du document, en mettant en avant les points principaux, les relations essentielles, les concepts , les dates et les lieux, les conclusions et les détails importants.
3. **Notes** :
- Présentez les points clés sous forme de liste à puces avec des émojis pertinents pour souligner la nature de chaque point.
- N'oubliez pas de relever tout les entités et les relations.
- Incluez des sous-points (sans émojis) sous les points principaux pour offrir des détails ou explications supplémentaires.
4. **Actions** : Identifiez et listez les actions spécifiques, tâches ou étapes recommandées ou nécessaires selon le contenu du document.
**Document :**
{text}
**Format de sortie :**
### Résumé :
[Fournissez un résumé concis du document ici;n'oubliez pas de relever tout les entités et les relations.]
### Notes :
- 📌 **Point Principal 1**
- Sous-point A
- Sous-point B
- 📈 **Point Principal 2**
- Sous-point C
- Sous-point D
- 📝 **Point Principal 3**
- Sous-point E
- Sous-point F
### Actions :
1. [Action 1]
2. [Action 2]
3. [Action 3]
4. ...
---
''')
if choice == "compte_rendu":
if "cr" not in st.session_state or st.session_state.cr == "":
with st.spinner("Génération du compte rendu..."):
cr = generate_response_via_langchain(prompt_cr,stream=False,model="gpt-4o")
st.session_state.cr = cr
st.session_state.cr_chat_history = []
else:
cr = st.session_state.cr
if cr:
col1, col2 = st.columns([2.5, 1.5])
with col1.container(border=True,height=850):
st.markdown("##### Compte rendu")
keywords_paragraph = f"### Mots clés extraits:\n- {audit['Mots clés'].strip()}"
with st.container(height=650,border=False):
st.markdown(keywords_paragraph)
st.write(cr)
# col_copy , col_success = st.columns([1,11])
# if col_copy.button("📋",key="copy_cr"):
with st.container(height=50,border=False):
st_copy_to_clipboard(keywords_paragraph+"\n\n"+cr,key="cp_but_cr")
# col_success.success("Compte rendu copié dans le presse-papier")
with col2.container(border=True,height=850):
st.markdown("##### Dialoguer avec le CR")
user_query = st.chat_input("Par ici ...")
if user_query is not None and user_query != "":
st.session_state.cr_chat_history.append(HumanMessage(content=user_query))
with st.container(height=600, border=False):
for message in st.session_state.cr_chat_history:
if isinstance(message, AIMessage):
with st.chat_message("AI"):
st.markdown(message.content)
elif isinstance(message, HumanMessage):
with st.chat_message("Human"):
st.write(message.content)
#check if last message is human message
if len(st.session_state.cr_chat_history) > 0:
last_message = st.session_state.cr_chat_history[-1]
if isinstance(last_message, HumanMessage):
with st.chat_message("AI"):
retreive = st.session_state.vectorstore.as_retriever()
context = retreive.invoke(last_message.content)
wrapped_prompt = f'''Étant donné le contexte suivant {context} et le compte rendu du document {cr}, {last_message.content}'''
response = st.write_stream(generate_response_via_langchain(wrapped_prompt,stream=True))
st.session_state.cr_chat_history.append(AIMessage(content=response))
# col_copy_c , col_success_c = st.columns([1,7])
# if col_copy_c.button("📋",key="copy_cr_chat"):
with st.container(height=50,border=False):
chat_formatted = chat_history_formatter(st.session_state.cr_chat_history)
st_copy_to_clipboard(chat_formatted,key="cp_but_cr_chat",show_text=False)
# col_success_c.success("Historique copié !")
elif choice == "graphe de connaissance":
if "graph" not in st.session_state or st.session_state.graph == None:
with st.spinner("Génération du graphe..."):
keywords_list = [keyword.strip() for keyword in audit["Mots clés"].strip().split(",")]
allowed_nodes_types =keywords_list+ ["Person","Organization","Location","Event","Date","Time","Ressource","Concept"]
number_tokens = audit["Nombre de tokens"]
if number_tokens > 10000:
if st.session_state.cr == "":
st.session_state.cr = generate_response_via_langchain(prompt_cr,stream=False,model="gpt-4o")
text = st.session_state.cr
graph = get_graph(text,allowed_nodes=allowed_nodes_types)
st.session_state.graph = graph
st.session_state.filter_views = {}
st.session_state.current_view = None
st.session_state.node_types = None
st.session_state.chat_graph_history = []
node_types = get_node_types(graph[0])
list_node_types = list(node_types)
sorted_node_types = sorted(list_node_types,key=lambda x: x.lower())
print(sorted_node_types)
nodes_type_dict = list_to_dict_colors(sorted_node_types)
st.session_state.node_types = nodes_type_dict
st.session_state.filter_views["Vue par défaut"] = list(node_types)
st.session_state.current_view = "Vue par défaut"
else:
graph = st.session_state.graph
if graph is not None:
#st.write(graph)
edges,nodes,config = convert_neo4j_to_agraph(graph[0],st.session_state.node_types)
col1, col2 = st.columns([2.5, 1.5])
with col1.container(border=True,height=850):
st.write("##### Visualisation du graphe (**"+st.session_state.current_view+"**)")
filter_col,add_view_col,change_view_col,color_col = st.columns([9,1,1,1])
if color_col.button("🎨",help="Changer la couleur"):
change_color_dialog()
if change_view_col.button("🔍",help="Changer de vue"):
change_view_dialog()
#add mots cles to evry label in audit["Mots clés"]
#filter_labels = [ label + " (mot clé)" if label.strip().lower() in audit["Mots clés"].strip().lower().split(",") else label for label in st.session_state.filter_views[st.session_state.current_view] ]
keywords_list = [keyword.strip().lower() for keyword in audit["Mots clés"].strip().split(",")]
dict_filters = {label: "Mot clé : "+label if label.strip().lower() in keywords_list else label for label in st.session_state.filter_views[st.session_state.current_view]}
default_target_filter = filter_correspondance(st.session_state.filter_views[st.session_state.current_view],dict_filters)
# st.write(default_target_filter)
# st.write(dict_filters)
sorted_default_target_filter = sorted(default_target_filter,key=lambda x: x.lower())
target_filter = filter_correspondance(list(st.session_state.node_types.keys()),dict_filters)
target_filter = sorted(target_filter,key=lambda x: x.lower())
filter = filter_col.multiselect("Filtrer selon l'étiquette",target_filter,placeholder="Sélectionner une ou plusieurs étiquettes",default=default_target_filter,label_visibility="collapsed")
filter = filter_correspondance(filter,dict_filters,reverse=True)
if add_view_col.button("➕",help="Ajouter une vue"):
add_view_dialog(filter)
if filter:
nodes = filter_nodes_by_types(nodes,filter)
selected = display_graph(edges,nodes,config)
# col_copy , col_success = st.columns([1,11])
# if col_copy.button("📋",key="copy_graph"):
with st.container(height=50,border=False):
graph_json = graph_doc_to_json(graph[0])
st_copy_to_clipboard(json.dumps(graph_json),key="cp_but_graph")
# col_success.success("Graphe copié dans le presse-papier")
with col2.container(border=True,height=850):
st.markdown("##### Dialoguer avec le graphe")
user_query = st.chat_input("Par ici ...")
if user_query is not None and user_query != "":
st.session_state.chat_graph_history.append(HumanMessage(content=user_query))
with st.container(height=600, border=False):
for message in st.session_state.chat_graph_history:
if isinstance(message, AIMessage):
with st.chat_message("AI"):
st.markdown(message.content)
elif isinstance(message, HumanMessage):
with st.chat_message("Human"):
st.write(message.content)
#check if last message is human message
if len(st.session_state.chat_graph_history) > 0:
last_message = st.session_state.chat_graph_history[-1]
if isinstance(last_message, HumanMessage):
with st.chat_message("AI"):
retreive = st.session_state.vectorstore.as_retriever()
context = retreive.invoke(last_message.content)
wrapped_prompt = f"Étant donné le contexte suivant {context}, et le graph de connaissance: {graph}, {last_message.content}"
response = st.write_stream(generate_response_via_langchain(wrapped_prompt,stream=True))
st.session_state.chat_graph_history.append(AIMessage(content=response))
if selected is not None:
with st.chat_message("AI"):
st.markdown(f" EXPLORER LES DONNEES CONTENUES DANS **{selected}**")
prompts = [f"Extrait moi toutes les informations du noeud ''{selected}'' ➡️",
f"Montre moi les conversations autour du noeud ''{selected}'' ➡️"]
for i,prompt in enumerate(prompts):
button = st.button(prompt,key=f"p_{i}",on_click=lambda i=i: st.session_state.chat_graph_history.append(HumanMessage(content=prompts[i])))
# col_copy_c , col_success_c = st.columns([1,7])
# if col_copy_c.button("📋",key="copy_graph_chat"):
with st.container(height=50,border=False):
st_copy_to_clipboard(chat_history_formatter(st.session_state.chat_graph_history),key="cp_but_graph_chat",show_text=False)
# col_success_c.success("Historique copié !")
doc_dialog_main()
|