通过检索增强生成(RAG) 增强LLM的实战演练

Shell 
1 (venv) ~/project$ pip install llama-index langchain langchain-openai
1.
2.

Shell 
1(venv) ~/project$ mkdir data
2
3 (venv) ~/project$ curl \
4           https://www.gutenberg.org/cache/epub/3296/pg3296.txt \
5              -o data/confessions.txt
6
7
8 (venv) ~/project$ ls data
9  confessions.txt
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.

Python 
1 # index.py
2
3 from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
4 import os
5
6 if os.environ.get('OPENAI_API_KEY') is None:
7  exit('You must provide an OPENAI_API_KEY env var.')
8
9 documents = SimpleDirectoryReader("data").load_data()
10 index = VectorStoreIndex.from_documents(documents)
11
12 query_engine = index.as_query_engine()
13 response = query_engine.query("In which city is Saint Augustine the Bishop?")
14 print(response)
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.

Shell 
1 (venv) ~/project$ OPENAI_API_KEY=sk-******** python index.py
2 Hippo
1.
2.
3.

Shell 
1 THE CONFESSIONS OF SAINT AUGUSTINE
2
3 By Saint Augustine
4
5 Bishop of Hippo
1.
2.
3.
4.
5.
6.

Python 
1 PERSIST_DIR='./my_vector_indexes/gutenberg/'
2 index.storage_context.persist(persist_dir=PERSIST_DIR)
1.
2.
3.

Shell 
1 (venv) ~/project$ OPENAI_API_KEY=sk-******** python index.py
2 Hippo
3
4 (venv) ~/project$ tree
5 .
6├── data
7│   └── confessions.txt
8├── index.py
9└── my_vector_indexes
10    └── gutenberg
11        ├── default__vector_store.json
12        ├── docstore.json
13        ├── graph_store.json
14        ├── image__vector_store.json
15        └── index_store.json
16
17 3 directories, 7 files
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.

Python 
1 # index.py
2
3 import os
4 from langchain_openai import ChatOpenAI
5 from langchain.chains import ConversationalRetrievalChain
6 from langchain.memory import ConversationBufferWindowMemory
7 from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
8 from langchain_community.retrievers import LlamaIndexRetriever
9 from fastapi import FastAPI
10 from pydantic import BaseModel
11
12 if os.environ.get('OPENAI_API_KEY') is None:
13  exit('You must provide an OPENAI_API_KEY env var.')
14
15 documents = SimpleDirectoryReader("data").load_data()
16 index = VectorStoreIndex.from_documents(documents)
17
18 # For this demo, we will not persist the index.
19
20
21 retriever = LlamaIndexRetriever(index=index.as_query_engine())
22
23 llm = ChatOpenAI(model_name="gpt-3.5-turbo", max_tokens=2048)
24
25 memory = ConversationBufferWindowMemory(
26  memory_key='chat_history',
27  return_messages=True,
28  k=3
29 )
30
31 conversation = ConversationalRetrievalChain.from_llm(
32  llm=llm, 
33  retriever=retriever,
34  memory=memory,
35  max_tokens_limit=1536  
36 )
37
38 class Prompt(BaseModel):
39  question: str
40
41 app = FastAPI()
42
43 @app.post("/prompt")
44 async def query_chatbot(prompt: Prompt):
45  response = conversation.invoke({'question': prompt.question})
46  return response['answer']
47
48 if __name__=='__main__':
49  import uvicorn
50  uvicorn.run(app, host="localhost", port=8000)
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.

Shell 
1(venv) ~/project$ pip install fastapi pydantic uvicorn
1.
2.

Shell 
1 (venv) ~/project$ OPENAI_API_KEY=sk-******** python index.py
2INFO:     Started server process [1101807]
3 INFO:     Waiting for application startup.
4 INFO:     Application startup complete.
5 INFO:     Uvicorn running on http://localhost:8000 (Press CTRL+C to quit)
1.
2.
3.
4.
5.
6.

Shell 
1 $ curl -X POST \
2       --header "Content-type:application/json" \
3       --data '{"question":"Who is Ambrose?"}' \
4       http://localhost:8000/prompt
5
6
7 "Ambrose is a person mentioned in the text provided. He is described as a respected
8 and celibate man who was esteemed by the author. Ambrose is depicted as a figure of
9 great honor and excellence, particularly known for his dedication to reading and
10 studying."
11
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.

Shell 
1(venv) ~/project$ pip freeze > requirements.txt
1.
2.

Shell 
1 (venv) ~/project$ echo \
2                 'web: uvicorn index:app --host=0.0.0.0 --port=${PORT}' > Procfile
1.
2.
3.

Shell 
1 (venv) ~/project$ echo 'python-3.11.8' > runtime.txt
1.
2.

Shell 
1 ~/project$ tree
2 .
3├── data
4│   └── confessions.txt
5├── index.py
6├── Procfile
7├── requirements.txt
8└── runtime.txt
9
10 1 directory, 5 files
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.

Shell 
1 ~/project$ heroku login
2
3 ~/project$ heroku apps:create my-llamaindex-app
4
5 ~/project$ heroku git:remote -a my-llamaindex-app
6
7 ~/project$ heroku config:add OPENAI_API_KEY=replaceme -a my-llamaindex-app
8
9 ~/project$ git push heroku main
10 …
11 remote: -----> Building on the Heroku-22 stack
12 remote: -----> Determining which buildpack to use for this app
13 remote: -----> Python app detected
14 remote: -----> Using Python version specified in runtime.txt
15 …
16 remote: -----> Launching...
17 remote:        Released v4
18 remote:        https://my-llamaindex-app-6b48faa3ee6a.herokuapp.com/ deployed to Heroku
19
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.

Shell 
1 $ curl -X POST \
2       --header "Content-type:application/json" \
3       --data '{"question":"Who is Ambrose?"}' \
4       https://my-llamaindex-app-6b48faa3ee6a.herokuapp.com/prompt
5
6 "Ambrose is a significant figure in the text provided. He is being described as a
7 respected and happy man, known for his celibacy and his dedication to reading and
8 studying. He is referred to as a holy oracle and a person of great influence and
9 wisdom."
10
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.