\"outputFieldMappings\": [\n    {\n      \"sourceFieldName\": \"/document/content/pages/*/keyphrases/*\",\n      \"targetFieldName\": \"keyphrases\"\n    },\n    {\n      \"sourceFieldName\": \"/document/content/pages/*\",\n      \"targetFieldName\": \"pages\"\n    }\n]

3. Set up the project and install the libraries

mkdir gpt-proj\ncd gpt-proj1

pip install openai==1.14.1
pip install langchain==0.1.11
pip install langchain_openai==0.0.8
pip install faiss-cpu==1.8.0

 4. Set up the project in VS Code

\n

1. In VS Code, select the folder that you have created.

\n
\n

- Open VS Code and select File > Open Folder from the menu. Select the gpt-proj1 folder that you created earlier, which is located at C:\\Users\\yourUserName\\azure-proj\\gpt-proj1.

\n

 

\n

\n

 

\n

2. Create a new file.

\n
\n

- In the left pane of VS Code, right-click and select New File to create a new file named example.py.

\n

 

\n

\n

 

\n

3. Import the required packages.

\n
\n

- Type the following code in the example.py file in VS Code.

\n

# Library imports\nfrom collections import OrderedDict\nimport requests\n\n# Lang Chain library imports\nfrom langchain.chains import RetrievalQAWithSourcesChain\nfrom langchain_openai import AzureOpenAIEmbeddings\nfrom langchain_openai.chat_models import AzureChatOpenAI\nfrom langchain.docstore.document import Document\nfrom langchain_community.vectorstores import FAISS\n

# Azure AI Search service settings\nSEARCH_SERVICE_NAME = 'your-search-service-name' # 'teachchatgpt-search'\nSEARCH_SERVICE_ENDPOINT = f'https://{SEARCH_SERVICE_NAME.lower()}.search.windows.net/'\nSEARCH_SERVICE_KEY = 'your-search-service-key'\nSEARCH_SERVICE_API_VERSION = 'your-API-version' # '2023-10-01-Preview'\n\n# Azure AI Search service index settings\nSEARCH_SERVICE_INDEX_NAME1 = 'your-search-service-index-name' # 'teachchatgpt-index'\n\n# Azure AI Search service semantic configuration settings\nSEARCH_SERVICE_SEMANTIC_CONFIG_NAME = 'your-semantic-configuration-name' # 'teachchatgpt-config'\n\n# Azure OpenAI settings\nAZURE_OPENAI_NAME = 'your-openai-name' # 'teachchatgpt-azureopenai'\nAZURE_OPENAI_ENDPOINT = f'https://{AZURE_OPENAI_NAME.lower()}.openai.azure.com/'\nAZURE_OPENAI_KEY = 'your-openai-key'\nAZURE_OPENAI_API_VERSION = 'your-API-version' # '2024-02-15-preview'

\n

5. Fill in your config.py file with your Azure information.

\n
\n

NOTE:

\n

You'll need to include information about your Azure AI Search Service name, index name, semantic configuration name, key, and API version, and Azure OpenAI name, key, and API version.

\n
\n

TIP:

\n

Find your Azure information

\n

1. Find the Azure AI Search Keys.

\n

 

\n

- Navigate to your AI Search service, then select Keys, then copy and paste your key into the config.py file.

\n

 

\n

\n

 

\n

2. Find the Azure AI Search Index name.

\n

 

\n

- Navigate to your AI Search service, then select Indexes, then copy and paste your index name into the config.py file.

\n

 

\n

\n

 

\n

3. Find the Azure AI Search Semantic configuration name.

\n

 

\n

- Navigate to your AI Search service, select Indexes, and then click your index name.

\n

- Select Semantic configurations and copy and paste your Semantic configuration name into the config.py file.

\n

 

\n

\n

 

\n

4. Find the Azure OpenAI Keys.

\n

 

\n

- Navigate to your Azure OpenAI, then select Keys and Endpoint, then copy and paste your key into the config.py file

\n

 

\n

\n

 

\n

5. Choose your Azure AI Search API and Azure OpenAI version.

\n

 

\n

- Select your version of the Azure AI Search API and Azure OpenAI API using the hyperlinks below.

\n

- Select an Azure AI Search API version 

\n

- Select an Azure OpenAI API version 

\n

- I have selected the latest version of the Azure AI Search API, 2023-10-01-Preview, and the Azure OpenAI API, 2024-02-15-Preview.

\n

 

\n

5. Search with Azure AI Search

\n

 

\n

\n

\n

In this section, we'll use Azure Cognitive Search within VS Code. We have already installed all the necessary packages in the previous chapter. Now we will focus on how to use Azure AI Search and Azure OpenAI in VS Code.

\n

In Chapters 5 and 6, we'll create functions that use Azure AI Service and Azure OpenAI and use them in example.py.

\n

To use Azure AI Search and Azure OpenAI, we need to import the information from Azure that we entered in config.py into example.py that we created earlier.

\n

All the following code comes from example.py.

\n

The full code is provided at the end of the chapter for your convenience.

\n

 

\n

1. Add code to example.py that imports the values from config.py.

\n

\n

 

\n

# Configuration imports\nfrom config import (\n    SEARCH_SERVICE_ENDPOINT,\n    SEARCH_SERVICE_KEY,\n    SEARCH_SERVICE_API_VERSION,\n    SEARCH_SERVICE_INDEX_NAME1,\n    SEARCH_SERVICE_SEMANTIC_CONFIG_NAME,\n    AZURE_OPENAI_ENDPOINT,\n    AZURE_OPENAI_KEY,\n    AZURE_OPENAI_API_VERSION,\n)

\n

 

\n

\n

2. Add the Azure AI Search Service header.

\n

\n

 

\n

# Azure AI Search service header settings\nHEADERS = {\n    'Content-Type': 'application/json',\n    'api-key': SEARCH_SERVICE_KEY\n}

\n

\n

3. Now, we will create functions related to Azure AI Search and run them from the main function.

\n
\n

-  Add the following code to the example.py file to create the `search_documents` function that retrieves documents related to your question.

\n

\n

 

\n

\ndef search_documents(question):\n    \"\"\"Search documents using Azure AI Search.\"\"\"\n    # Construct the Azure AI Search service access URL.\n    url = (SEARCH_SERVICE_ENDPOINT + 'indexes/' +\n                SEARCH_SERVICE_INDEX_NAME1 + '/docs')\n    \n    # Create a parameter dictionary.\n    params = {\n        'api-version': SEARCH_SERVICE_API_VERSION,\n        'search': question,\n        'select': '*',\n        # '$top': 5, Extract the top 5 documents from your storage.\n        '$top': 5,\n        'queryLanguage': 'en-us',\n        'queryType': 'semantic',\n        'semanticConfiguration': SEARCH_SERVICE_SEMANTIC_CONFIG_NAME,\n        '$count': 'true',\n        'speller': 'lexicon',\n        'answers': 'extractive|count-3',\n        'captions': 'extractive|highlight-false'\n        }\n    # Make a GET request to the Azure AI Search service and store the response in a variable.\n    resp = requests.get(url, headers=HEADERS, params=params)\n    # Return the JSON response containing the search results.\n    search_results = resp.json()\n\n    return search_results

\n

 

\n

\n

\n

- Add the following code to the example.py file to create the `filter_documents` function that extracts documents with a certain score or higher based on how relevant it is to your question.

\n

 

\n

def filter_documents(search_results):\n    \"\"\"Filter documents with a reranker score above a certain threshold.\"\"\"\n    documents = OrderedDict()\n    for result in search_results['value']:\n        # The '@search.rerankerScore' range is 0 to 4.00, where a higher score indicates a stronger semantic match.\n        if result['@search.rerankerScore'] > 0.8:\n            documents[result['metadata_storage_path']] = {\n                'chunks': result['pages'][:10],\n                'captions': result['@search.captions'][:10],\n                'score': result['@search.rerankerScore'],\n                'file_name': result['metadata_storage_name']\n            }\n\n    return documents

\n

\n

\n

\n

 

\n

4. Now we'll run the code using the above functions in the main function.

\n
\n

- Add the following code to the example.py file to create the `main` function.

\n

- When you run it, you'll see the total number of PDFs in the blob storage, the top few documents adopted, and the number of chunks.

\n

- I asked the question, 'Tell me about effective prompting strategies' based on the paper I had stored on the blob storage.

\n

- If you want to see the full search results, add `print(search_results)` to your main function.

\n

\n

\n

def main():\n\n    QUESTION = 'Tell me about effective prompting strategies'\n\n    # Search for documents with Azure AI Search\n\n    search_results = search_documents(QUESTION)\n\n    documents = filter_documents(search_results)\n\n    print('Total Documents Found: {}, Top Documents: {}'.format(\n        search_results['@odata.count'], len(search_results['value'])))\n\n\n    # 'chunks' is the value that corresponds to the Pages field that you set up in the AI Search service.\n    docs = []\n    for key,value in file_content.items():\n        for page in value['chunks']:\n            docs.append(Document(page_content = page,\n                                metadata={\"source\": value[\"file_name\"]}))\n\n# execute the main function\nif __name__ == \"__main__\":\n    main()

6. Get answers from PDF content using Azure OpenAI and AI Search

Now that Azure AI Search is working well in VS Code, it's time to start using
Azure OpenAI.
In this chapter, we'll create functions related to Azure OpenAI and ultimately create
and run a program in example.py file that answers a question with Azure OpenAI based on
the search information from Azure AI Search.

1. We will create functions related to Azure OpenAI and Lang Chain and run them from
the main function.

- Add the following code to the example.py file to create the `create_embeddings` function that creates an embedding model.

def create_embeddings():\n    \"\"\"Create an embedding model.\"\"\"\n    embeddings = AzureOpenAIEmbeddings(\n        openai_api_key = AZURE_OPENAI_KEY,\n        azure_endpoint = AZURE_OPENAI_ENDPOINT,\n        openai_api_version = AZURE_OPENAI_API_VERSION,\n        openai_api_type = 'azure',\n        azure_deployment = 'text-embedding-ada-002',\n        model = 'text-embedding-ada-002',\n        chunk_size=1\n    )\n    return embeddings\n

- Add the following code to the example.py file to create the `store_documents` function that stores the extracted documents in a vector database.

\ndef store_documents(docs, embeddings):\n    \"\"\"Create vector store and store documents in the vector store.\"\"\"\n    return FAISS.from_documents(docs, embeddings)\n

-  Add the following code to the example.py file to create the `answer_with_langchain` function that extracts the chunk associated with a question stored in the vector database and answers the question based on it.

def answer_with_langchain(vector_store, question):\n    \"\"\"Search for documents related to your question from the vector store\n    and answer question with search result using Lang Chain.\"\"\"\n\n    # Add a chat service.\n    llm = AzureChatOpenAI(\n        openai_api_key = AZURE_OPENAI_KEY,\n        azure_endpoint = AZURE_OPENAI_ENDPOINT,\n        openai_api_version= AZURE_OPENAI_API_VERSION,\n        azure_deployment = 'gpt-35-turbo',\n        temperature=0.0,\n        max_tokens=500\n    )\n\n    chain = RetrievalQAWithSourcesChain.from_chain_type(\n        llm=llm,\n        chain_type='stuff',\n        retriever=vector_store.as_retriever(),\n        return_source_documents=True\n    )\n\n    answer = chain.invoke({'question': question})\n\n    return answer

2. Modify the `main` function to match this code.

def main():\n\n    QUESTION = 'Tell me about effective prompting strategies'\n\n    # Search for documents with Azure AI Search.\n\n    search_results = search_documents(QUESTION)\n\n    documents = filter_documents(search_results)\n\n    print('Total Documents Found: {}, Top Documents: {}'.format(\n        search_results['@odata.count'], len(search_results['value'])))\n\n\n    # 'chunks' is the value that corresponds to the Pages field that you set up in the AI Search service.\n    docs = []\n    for key,value in documents.items():\n        for page in value['chunks']:\n            docs.append(Document(page_content = page,\n                                metadata={\"source\": value[\"file_name\"]}))\n\n    # Answer your question using Lang Chain.\n\n    embeddings = create_embeddings()\n\n    vector_store = store_documents(docs, embeddings)\n\n    result = answer_with_langchain(vector_store, QUESTION)\n\n    print('Question: ', QUESTION)\n    print('Answer: ', result['answer'])\n    print('Reference: ', result['sources'].replace(\",\",\"\\n\"))

3. Now let's run it and see if it answers your question.
- The result of executing the code.

Total Documents Found: 5, Top Documents: 3
Number of chunks: 10
Question: Tell me about effective prompting strategies
Answer: Effective prompting strategies for improving the reliability of GPT-3 include 
establishing simple prompts that improve GPT-3's reliability in terms of generalizability, 
social biases, calibration, and factuality. These strategies include prompting with 
randomly sampled examples from the source domain, using examples sampled from 
a balanced demographic distribution and natural language intervention to reduce 
social biases, calibrating output probabilities, and updating the LLM's factual 
knowledge and reasoning chains. Natural language intervention can also effectively 
guide model predictions towards better fairness. 
Reference: Prompting GPT-3 To Be Reliable.pdf

# Azure AI Search service settings\nSEARCH_SERVICE_NAME = 'teachchatgpt-search' # 'teachchatgpt-search'\nSEARCH_SERVICE_ENDPOINT = f'https://{SEARCH_SERVICE_NAME.lower()}.search.windows.net/'\nSEARCH_SERVICE_KEY = 'lDdgBQSnG4f0uAZixTMgr9nSOa4VvmeVux4Nac9iQXAzSeBm7ToL'\nSEARCH_SERVICE_API_VERSION = '2023-10-01-Preview' # '2023-10-01-Preview'\n\n# Azure AI Search service index settings\nSEARCH_SERVICE_INDEX_NAME1 = 'teachchatgpt-index' # 'teachchatgpt-index'\n\n# Azure AI Search service semantic configuration settings\nSEARCH_SERVICE_SEMANTIC_CONFIG_NAME = 'teachchatgpt-config' # 'teachchatgpt-config'\n\n# Azure OpenAI settings\nAZURE_OPENAI_NAME = 'teachchatgpt-azureopenai' # 'teachchatgpt-azureopenai'\nAZURE_OPENAI_ENDPOINT = f'https://{AZURE_OPENAI_NAME.lower()}.openai.azure.com/'\nAZURE_OPENAI_KEY = 'fdfcd358b26343c293d3dff4643f1aee'\nAZURE_OPENAI_API_VERSION = '2024-02-15-preview' # '2024-02-15-preview'\n

# Library imports\nfrom collections import OrderedDict\nimport requests\n\n# Lang Chain library imports\nfrom langchain.chains import RetrievalQAWithSourcesChain\nfrom langchain_openai import AzureOpenAIEmbeddings\nfrom langchain_openai.chat_models import AzureChatOpenAI\nfrom langchain.docstore.document import Document\nfrom langchain_community.vectorstores import FAISS\n\n# Configuration imports\nfrom config import (\n    SEARCH_SERVICE_ENDPOINT,\n    SEARCH_SERVICE_KEY,\n    SEARCH_SERVICE_API_VERSION,\n    SEARCH_SERVICE_INDEX_NAME1,\n    SEARCH_SERVICE_SEMANTIC_CONFIG_NAME,\n    AZURE_OPENAI_ENDPOINT,\n    AZURE_OPENAI_KEY,\n    AZURE_OPENAI_API_VERSION,\n)\n\n# Azure AI Search service header settings\nHEADERS = {\n    'Content-Type': 'application/json',\n    'api-key': SEARCH_SERVICE_KEY\n}\n\ndef search_documents(question):\n    \"\"\"Search documents using Azure AI Search.\"\"\"\n    # Construct the Azure AI Search service access URL.\n    url = (SEARCH_SERVICE_ENDPOINT + 'indexes/' +\n                SEARCH_SERVICE_INDEX_NAME1 + '/docs')\n    \n    # Create a parameter dictionary.\n    params = {\n        'api-version': SEARCH_SERVICE_API_VERSION,\n        'search': question,\n        'select': '*',\n        # '$top': 5, Extract the top 5 documents from your storage.\n        '$top': 5,\n        'queryLanguage': 'en-us',\n        'queryType': 'semantic',\n        'semanticConfiguration': SEARCH_SERVICE_SEMANTIC_CONFIG_NAME,\n        '$count': 'true',\n        'speller': 'lexicon',\n        'answers': 'extractive|count-3',\n        'captions': 'extractive|highlight-false'\n        }\n    # Make a GET request to the Azure AI Search service and store the response in a variable.\n    resp = requests.get(url, headers=HEADERS, params=params)\n    # Return the JSON response containing the search results.\n    search_results = resp.json()\n\n    return search_results\n\ndef filter_documents(search_results):\n    \"\"\"Filter documents with a reranker score above a certain threshold.\"\"\"\n    documents = OrderedDict()\n    for result in search_results['value']:\n        # The '@search.rerankerScore' range is 0 to 4.00, where a higher score indicates a stronger semantic match.\n        if result['@search.rerankerScore'] > 0.8:\n            documents[result['metadata_storage_path']] = {\n                'chunks': result['pages'][:10],\n                'captions': result['@search.captions'][:10],\n                'score': result['@search.rerankerScore'],\n                'file_name': result['metadata_storage_name']\n            }\n\n    return documents\n\ndef create_embeddings():\n    \"\"\"Create an embedding model.\"\"\"\n    embeddings = AzureOpenAIEmbeddings(\n        openai_api_key = AZURE_OPENAI_KEY,\n        azure_endpoint = AZURE_OPENAI_ENDPOINT,\n        openai_api_version = AZURE_OPENAI_API_VERSION,\n        openai_api_type = 'azure',\n        azure_deployment = 'text-embedding-ada-002',\n        model = 'text-embedding-ada-002',\n        chunk_size=1\n    )\n    return embeddings\n\ndef store_documents(docs, embeddings):\n    \"\"\"Create vector store and store documents in the vector store.\"\"\"\n    return FAISS.from_documents(docs, embeddings)\n\ndef answer_with_langchain(vector_store, question):\n    \"\"\"Search for documents related to your question from the vector store\n    and answer question with search result using Lang Chain.\"\"\"\n\n    # Add a chat service.\n    llm = AzureChatOpenAI(\n        openai_api_key = AZURE_OPENAI_KEY,\n        azure_endpoint = AZURE_OPENAI_ENDPOINT,\n        openai_api_version= AZURE_OPENAI_API_VERSION,\n        azure_deployment = 'gpt-35-turbo',\n        temperature=0.0,\n        max_tokens=500\n    )\n\n    chain = RetrievalQAWithSourcesChain.from_chain_type(\n        llm=llm,\n        chain_type='stuff',\n        retriever=vector_store.as_retriever(),\n        return_source_documents=True\n    )\n\n    answer = chain.invoke({'question': question})\n\n    return answer\n\ndef main():\n    QUESTION = 'Tell me about effective prompting strategies'\n\n    # Search for documents with Azure AI Search.\n\n    search_results = search_documents(QUESTION)\n\n    documents = filter_documents(search_results)\n\n    print('Total Documents Found: {}, Top Documents: {}'.format(\n        search_results['@odata.count'], len(search_results['value'])))\n\n    \n    # 'chunks' is the value that corresponds to the Pages field that you set up in the AI Search service.\n    docs = []\n    for key,value in documents.items():\n        for page in value['chunks']:\n            docs.append(Document(page_content = page,\n                                metadata={\"source\": value[\"file_name\"]}))\n\n    # Answer your question using Lang Chain.\n\n    embeddings = create_embeddings()\n\n    vector_store = store_documents(docs, embeddings)\n\n    result = answer_with_langchain(vector_store, QUESTION)\n\n    print('Question: ', QUESTION)\n    print('Answer: ', result['answer'])\n    print('Reference: ', result['sources'].replace(\",\",\"\\n\"))\n\n# Execute the main function.\nif __name__ == \"__main__\":\n    main()

Next Steps

Documentation

Training Content

3. Set up the project and install the libraries

mkdir gpt-proj\ncd gpt-proj1

pip install openai==1.14.1
pip install langchain==0.1.11
pip install langchain_openai==0.0.8
pip install faiss-cpu==1.8.0

 4. Set up the project in VS Code

6. Get answers from PDF content using Azure OpenAI and AI Search

Now that Azure AI Search is working well in VS Code, it's time to start using
Azure OpenAI.
In this chapter, we'll create functions related to Azure OpenAI and ultimately create
and run a program in example.py file that answers a question with Azure OpenAI based on
the search information from Azure AI Search.

1. We will create functions related to Azure OpenAI and Lang Chain and run them from
the main function.

- Add the following code to the example.py file to create the `create_embeddings` function that creates an embedding model.

- Add the following code to the example.py file to create the `store_documents` function that stores the extracted documents in a vector database.

-  Add the following code to the example.py file to create the `answer_with_langchain` function that extracts the chunk associated with a question stored in the vector database and answers the question based on it.

2. Modify the `main` function to match this code.

3. Now let's run it and see if it answers your question.
- The result of executing the code.

Total Documents Found: 5, Top Documents: 3
Number of chunks: 10
Question: Tell me about effective prompting strategies
Answer: Effective prompting strategies for improving the reliability of GPT-3 include 
establishing simple prompts that improve GPT-3's reliability in terms of generalizability, 
social biases, calibration, and factuality. These strategies include prompting with 
randomly sampled examples from the source domain, using examples sampled from 
a balanced demographic distribution and natural language intervention to reduce 
social biases, calibrating output probabilities, and updating the LLM's factual 
knowledge and reasoning chains. Natural language intervention can also effectively 
guide model predictions towards better fairness. 
Reference: Prompting GPT-3 To Be Reliable.pdf

Next Steps

Documentation

Training Content