fixed. #16

.env.example

@@ -33,6 +33,7 @@ E2B_API_KEY=""
 
 # for web ui
 DATA_DIR="~/.cache/chatpilot/data"
+AGENTICA_HOME="~/.cache/chatpilot/data"
 
 # DO NOT TRACK
 SCARF_NO_ANALYTICS=true

chatpilot/apps/rag_app.py

@@ -18,6 +18,11 @@
     Embeddings,
 )
 from chromadb.utils import embedding_functions
+from chromadb.utils.embedding_functions.openai_embedding_function import OpenAIEmbeddingFunction
+from chromadb.utils.embedding_functions.text2vec_embedding_function import Text2VecEmbeddingFunction
+from chromadb.utils.embedding_functions.sentence_transformer_embedding_function import (
+    SentenceTransformerEmbeddingFunction
+)
 from fastapi import (
     FastAPI,
     Depends,
@@ -116,13 +121,13 @@ def __call__(self, input: ChromaDocuments) -> Embeddings:
 
 
 if "text-embedding" in app.state.RAG_EMBEDDING_MODEL and app.state.OPENAI_API_KEYS and app.state.OPENAI_API_KEYS[0]:
-    app.state.sentence_transformer_ef = embedding_functions.OpenAIEmbeddingFunction(
+    app.state.sentence_transformer_ef = OpenAIEmbeddingFunction(
         api_key=app.state.OPENAI_API_KEYS[0],
         api_base=app.state.OPENAI_API_BASE_URLS[0],
         model_name=app.state.RAG_EMBEDDING_MODEL,
     )
 elif "text2vec" in app.state.RAG_EMBEDDING_MODEL:
-    app.state.sentence_transformer_ef = embedding_functions.Text2VecEmbeddingFunction(
+    app.state.sentence_transformer_ef = Text2VecEmbeddingFunction(
         model_name=app.state.RAG_EMBEDDING_MODEL
     )
 elif "w2v" in app.state.RAG_EMBEDDING_MODEL:
@@ -237,7 +242,7 @@ async def update_embedding_model(
             model_name=app.state.RAG_EMBEDDING_MODEL
         )
     elif app.state.RAG_EMBEDDING_MODEL:
-        app.state.sentence_transformer_ef = embedding_functions.SentenceTransformerEmbeddingFunction(
+        app.state.sentence_transformer_ef = SentenceTransformerEmbeddingFunction(
             model_name=app.state.RAG_EMBEDDING_MODEL
         )
     else:

chatpilot/langchain_assistant.py

@@ -31,6 +31,8 @@
     ENABLE_URL_CRAWLER_TOOL,
     ENABLE_RUN_PYTHON_CODE_TOOL,
     REACT_RPOMPT,
+    OPENAI_API_KEY,
+    OPENAI_API_BASE,
 )
 from chatpilot.react_parser import ReActParserAndNoTool
 
@@ -40,8 +42,8 @@ def __init__(
             self,
             model_type: str = "openai",
             model_name: str = "gpt-3.5-turbo-1106",
-            model_api_key: str = os.getenv("OPENAI_API_KEY"),
-            model_api_base: str = os.getenv("OPENAI_API_BASE"),
+            model_api_key: str = os.getenv("OPENAI_API_KEY") or OPENAI_API_KEY,
+            model_api_base: str = os.getenv("OPENAI_API_BASE") or OPENAI_API_BASE,
             search_name: Optional[str] = "serper",
             agent_type: str = "react",
             enable_search_tool: Optional[bool] = None,

chatpilot/rag_fusion.py

@@ -14,15 +14,19 @@
 from langchain_core.prompts import ChatPromptTemplate
 from langchain_core.runnables import RunnablePassthrough
 from langchain_openai import ChatOpenAI, OpenAIEmbeddings
+from langchain_community.embeddings.text2vec import Text2vecEmbeddings
+from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings
 from loguru import logger
-
-from chatpilot.config import RAG_TEMPLATE
+import os
+from chatpilot.config import RAG_TEMPLATE,OPENAI_API_KEY,OPENAI_API_BASE
 
 
 class RagFusion:
     def __init__(
             self,
             documents: List[Document],
+            model_api_key: str = os.getenv("OPENAI_API_KEY") or OPENAI_API_KEY,
+            model_api_base: str = os.getenv("OPENAI_API_BASE") or OPENAI_API_BASE,
             openai_model: str = "gpt-3.5-turbo-1106",
             generate_model: str = "gpt-3.5-turbo-16k",
             temperature: float = 0.0,
@@ -41,19 +45,26 @@ def __init__(
         ])
         # Using LLM generate more queries
         self.requery_model = ChatOpenAI(
+            openai_api_key=model_api_key,
+            openai_api_base=model_api_base,
             temperature=temperature,
             model=openai_model,
         )
         vectorstore = Chroma.from_documents(
             documents,
-            OpenAIEmbeddings()
+            OpenAIEmbeddings(
+                openai_api_key=model_api_key,
+                openai_api_base=model_api_base,
+            )
         )
         self.retriever = vectorstore.as_retriever()
 
         self.rag_prompt = ChatPromptTemplate.from_template(RAG_TEMPLATE)
 
         # LLM to RAG model
         self.generate_model = ChatOpenAI(
+            openai_api_key=model_api_key,
+            openai_api_base=model_api_base,
             temperature=temperature,
             model=generate_model,
         )
@@ -110,7 +121,8 @@ def run(self, question: str):
     }
 
     text_documents = [Document(page_content=doc, metadata={"source": f"{id}"}) for id, doc in all_documents.items()]
+    print('text_documents:', text_documents)
     rag_fusion = RagFusion(text_documents)
-    question = "气候变化的影响"
+    question = "气候变化及其经济影响"
     result = rag_fusion.run(question)
     print(result)

examples/chat_agent_demo.py

@@ -82,26 +82,3 @@ def demo5():
             print("===")
 
 
-    def demo6():
-        import asyncio
-        async def d():
-            m = LangchainAssistant()
-
-            questions = [
-                "俄罗斯今日新闻top3",
-                # "人体最大的器官是啥",
-                # "how many letters in the word 'educabe'?",
-                # "它是一个真的单词吗？",
-            ]
-            for i in questions:
-                print(i)
-                events = await m.astream_run(i)
-                async for event in events:
-                    print(event)
-                    print("===")
-                    pass
-
-        asyncio.run(d())
-
-
-    demo6()

examples/chat_demo.py

@@ -19,15 +19,13 @@
     def demo6():
         import asyncio
         m = LangchainAssistant(
-            model_type='azure',
-            model_name="gpt-35-turbo",
+            model_type='openai',
+            model_name="gpt-3.5-turbo",
             model_api_key=os.getenv("OPENAI_API_KEYS"),
             model_api_base=os.getenv("OPENAI_API_BASE_URLS"),
             search_name="serper",
             agent_type="react",
             enable_search_tool=True,
-            enable_run_python_code_tool=False,
-            enable_crawler_tool=False,
             streaming=True,
         )
         async def d():
@@ -51,19 +49,12 @@ async def d():
 
     def demo5():
         m = LangchainAssistant(
-            model_type='azure',
-            model_name="gpt-35-turbo",
-            model_api_key=os.getenv("OPENAI_API_KEYS"),
-            model_api_base=os.getenv("OPENAI_API_BASE_URLS"),
             search_name="serper",
             agent_type="react",
             enable_search_tool=True,
             enable_run_python_code_tool=True,
-            enable_crawler_tool=False,
         )
         questions = [
-            # "今天的俄罗斯相关的新闻top3有哪些？",
-            # "今天北京的天气怎么样？",
             "人类最大的器官是？"
         ]
         for i in questions:

examples/rag_demo.py

@@ -15,9 +15,11 @@
     file_url = "https://docs.smith.langchain.com/overview"
     loader = WebBaseLoader(file_url)
     docs = loader.load()
+    print(docs)
     web_documents = RecursiveCharacterTextSplitter(
-        chunk_size=1000, chunk_overlap=200
+        chunk_size=2000, chunk_overlap=200
     ).split_documents(docs)
+    print('web_documents:', web_documents)
     rag_fusion = RagFusion(web_documents)
     question = "LangSmith是啥"
     result = rag_fusion.run(question)

requirements.txt

@@ -2,6 +2,7 @@ jieba>=0.39
 loguru
 tqdm
 pandas
+tiktoken
 langchain~=0.1.11
 langchain-community~=0.0.27
 langchain-openai~=0.0.8