base de données v2

app.py

@@ -61,8 +61,8 @@ openai.api_version = "2022-12-01"
 
 retrieve_all = EmbeddingRetriever(
     document_store=FAISSDocumentStore.load(
-        index_path="./documents/climate_gpt.faiss",
-        config_path="./documents/climate_gpt.json",
+        index_path="./documents/climate_gpt_v2.faiss",
+        config_path="./documents/climate_gpt_v2.json",
     ),
     embedding_model="sentence-transformers/multi-qa-mpnet-base-dot-v1",
     model_format="sentence_transformers",
@@ -71,8 +71,8 @@ retrieve_all = EmbeddingRetriever(
 
 retrieve_giec = EmbeddingRetriever(
     document_store=FAISSDocumentStore.load(
-        index_path="./documents/climate_gpt_only_giec.faiss",
-        config_path="./documents/climate_gpt_only_giec.json",
+        index_path="./documents/climate_gpt_v2_only_giec.faiss",
+        config_path="./documents/climate_gpt_v2_only_giec.json",
     ),
     embedding_model="sentence-transformers/multi-qa-mpnet-base-dot-v1",
     model_format="sentence_transformers",
@@ -124,18 +124,17 @@ def chat(
         stop=["\n---\n", "<|im_end|>"],
     )
     reformulated_query = reformulated_query["choices"][0]["text"]
-    reformulated_query,language = reformulated_query.split("\n")
+    reformulated_query, language = reformulated_query.split("\n")
     language = language.split(":")[1].strip()
     docs = [d for d in retriever.retrieve(query=reformulated_query, top_k=10) if d.score > threshold]
     messages = history + [{"role": "user", "content": query}]
 
     if docs:
-        
         docs_string = []
-        for i,d in enumerate(docs,1):
-            content = d.content.replace("\r\n","")
+        for i, d in enumerate(docs, 1):
+            content = d.content.replace("\r\n", "")
             docs_string.append(f"📃 doc {i}: {d.meta['file_name']} page {d.meta['page_number']}\n{content}")
-        sources = "\n\n".join([f"Query used for retrieval:\n{reformulated_query}"]+docs_string)
+        sources = "\n\n".join([f"Query used for retrieval:\n{reformulated_query}"] + docs_string)
         messages.append({"role": "system", "content": f"{sources_prompt}\n\n{sources}\n\nAnswer in {language}:"})
 
         response = openai.Completion.create(

climate_gpt.db

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:38229336d47aa1c618a57443ccca94a13af559732b9f94754c00dd96563a1570
-size 105828352

climate_gpt_only_giec.db

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:2063cac42eab1bdebe38a4b7f9ee79545d4f4999d1288e637ffcc7701104ccdb
-size 76234752

documents/climate_gpt.faiss

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5f76824baf8ab91efd808728e6e4fbf31b6a30d621b3379835812ab0f75e9936
-size 178775085

documents/climate_gpt.json

@@ -1,7 +0,0 @@
-{
-    "sql_url": "sqlite:///climate_gpt.db",
-    "faiss_index_factory_str": "Flat",
-    "index": "climate_gpt",
-    "similarity": "dot_product",
-    "embedding_dim": 768
-}

documents/climate_gpt_only_giec.faiss

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:bd872dea502921a947d1339193b2587d6708efd7da4be5c49009bc95caacb356
-size 116155437

documents/climate_gpt_only_giec.json

@@ -1 +0,0 @@
-{"sql_url": "sqlite:///climate_gpt_only_giec.db", "faiss_index_factory_str": "Flat", "index": "climate_gpt_only_giec", "similarity": "dot_product", "embedding_dim": 768}