'commit'

dsLightRag/Backup/PG_Start.py

@@ -109,7 +109,7 @@ async def rag(request: fastapi.Request):
 
             resp = await rag.aquery(
                 query=query,
-                param=QueryParam(mode="hybrid", stream=True, user_prompt=user_prompt))
+                param=QueryParam(mode="hybrid", stream=True, user_prompt=user_prompt,enable_rerank=True))
             async for chunk in resp:
                 if not chunk:
                     continue

dsLightRag/Doc/2、Conda维护.txt

@@ -17,14 +17,5 @@ conda list
 # 激活虚拟环境
 conda activate py310
 
-# 对虚拟环境中安装额外的包
-conda install -n py310 $package_name
-
 # 删除虚拟环境
-conda remove -n py310 --all
-
-# 删除环境中的某个包
-conda remove --name py310  $package_name
-
-# 恢复默认镜像
-conda config --remove-key channels
+conda remove -n py310 --all

dsLightRag/Util/LightRagUtil.py

@@ -155,13 +155,38 @@ def create_embedding_func():
     )
 
 
-async def rerank_func(query: str, documents: list, top_k: int = None, **kwargs):
-    return await jina_rerank(
-        query=query,
-        documents=documents,
-        model=Config.Config.RERANK_MODEL,
-        base_url=Config.Config.RERANK_BASE_URL,
-        api_key=Config.Config.RERANK_BINDING_API_KEY,
-        top_k=top_k or 10,
-        **kwargs
-    )
+
+# AGE
+os.environ["POSTGRES_HOST"] = POSTGRES_HOST
+os.environ["POSTGRES_PORT"] = str(POSTGRES_PORT)
+os.environ["POSTGRES_USER"] = POSTGRES_USER
+os.environ["POSTGRES_PASSWORD"] = POSTGRES_PASSWORD
+os.environ["POSTGRES_DATABASE"] = POSTGRES_DATABASE
+
+
+async def initialize_pg_rag(WORKING_DIR, workspace='default'):
+    rag = LightRAG(
+        working_dir=WORKING_DIR,
+        llm_model_func=llm_model_func,
+        llm_model_name=LLM_MODEL_NAME,
+        llm_model_max_async=4,
+        llm_model_max_token_size=32768,
+        enable_llm_cache_for_entity_extract=True,
+        embedding_func=EmbeddingFunc(
+            embedding_dim=EMBED_DIM,
+            max_token_size=EMBED_MAX_TOKEN_SIZE,
+            func=embedding_func
+        ),
+        rerank_model_func=rerank_func,
+        kv_storage="PGKVStorage",
+        doc_status_storage="PGDocStatusStorage",
+        graph_storage="PGGraphStorage",
+        vector_storage="PGVectorStorage",
+        auto_manage_storages_states=False,
+        vector_db_storage_cls_kwargs={"workspace": workspace}
+    )
+
+    await rag.initialize_storages()
+    await initialize_pipeline_status()
+
+    return rag