feat: setting embedding url and model in .env

Author: yuhexiong

.env.example

@@ -1,4 +1,4 @@
-# OPENAI or OLLAMA
+# LLM OPENAI or OLLAMA
 LLM_TYPE='OLLAMA'
 
 OPENAI_MODEL='gpt-4o-mini'
@@ -7,4 +7,9 @@ OPENAI_API_KEY=''
 OLLAMA_URL='http://localhost:11434'
 OLLAMA_MODEL='gemma3:27b'
 
+# Embedding
+OLLAMA_EMBEDDING_URL='http://localhost:11434'
+OLLAMA_EMBEDDING_MODEL='llama3.1:70b-instruct-q2_K'
+
+# Database
 DB_URL='postgresql://user:password@host:5432/database'

README-CH.md

@@ -13,15 +13,20 @@
 可以選擇 OpenAI 或 Ollama 作為 LLM，並設置所選擇 LLM 的環境變數。如果使用 OpenAI，則需要從 [OpenAI](https://platform.openai.com/) 獲取 API 金鑰（可能需要付費訂閱）。
 
 ```
-# OPENAI 或 OLLAMA
+# LLM OPENAI or OLLAMA
 LLM_TYPE='OLLAMA'
 
 OPENAI_MODEL='gpt-4o-mini'
 OPENAI_API_KEY=''
 
 OLLAMA_URL='http://localhost:11434'
-OLLAMA_MODEL='llama3.1:70b-instruct-q2_K'
+OLLAMA_MODEL='gemma3:27b'
 
+# Embedding
+OLLAMA_EMBEDDING_URL='http://localhost:11434'
+OLLAMA_EMBEDDING_MODEL='llama3.1:70b-instruct-q2_K'
+
+# Database
 DB_URL='postgresql://user:password@host:5432/database'
 ```
 

README.md

@@ -14,15 +14,20 @@ Use LangChain to integrate with OpenAI or Ollama and connect to the database, en
 Able to choose OpenAI or Ollama as the LLM, and set the environment variables for the selected LLM. If using OpenAI, an API key must be obtained from [OpenAI](https://platform.openai.com/) (a paid subscription may be required).
 
 ```
-# OPENAI or OLLAMA
+# LLM OPENAI or OLLAMA
 LLM_TYPE='OLLAMA'
 
 OPENAI_MODEL='gpt-4o-mini'
 OPENAI_API_KEY=''
 
 OLLAMA_URL='http://localhost:11434'
-OLLAMA_MODEL='llama3.1:70b-instruct-q2_K'
+OLLAMA_MODEL='gemma3:27b'
 
+# Embedding
+OLLAMA_EMBEDDING_URL='http://localhost:11434'
+OLLAMA_EMBEDDING_MODEL='llama3.1:70b-instruct-q2_K'
+
+# Database
 DB_URL='postgresql://user:password@host:5432/database'
 ```
 

rag_util.py

@@ -20,12 +20,16 @@
 # 讀取 .env 變數
 load_dotenv()
 
-
 def get_vector_store():
     # embedding
-    # TODO: 環境變數讀取
-    OLLAMA_URL = os.getenv("OLLAMA_URL")
-    embeddings = OllamaEmbeddings(model="llama3.1:70b-instruct-q2_K", base_url=OLLAMA_URL)
+    OLLAMA_EMBEDDING_URL = os.getenv("OLLAMA_EMBEDDING_URL")
+    if not OLLAMA_EMBEDDING_URL:
+        raise Exception("未在 .env 檔案中找到 OLLAMA_EMBEDDING_URL。")
+    OLLAMA_EMBEDDING_MODEL = os.getenv("OLLAMA_EMBEDDING_MODEL")
+    if not OLLAMA_EMBEDDING_MODEL:
+        raise Exception("未在 .env 檔案中找到 OLLAMA_EMBEDDING_MODEL。")
+
+    embeddings = OllamaEmbeddings(model=OLLAMA_EMBEDDING_MODEL, base_url=OLLAMA_EMBEDDING_URL)
 
     vector_store = InMemoryVectorStore(embeddings)