context_str to the console, you can use the format method of the QuestionAnswerPrompt instance. Here's how you can do it:context_str = "The author grew up in a small town, playing sports and reading books."
query_str = "What did the author do growing up?"
QA_PROMPT_TMPL = (
"We have provided context information below. \n"
"---------------------\n"
"{context_str}"
"\n---------------------\n"
"Given this information, please answer the question: {query_str}\n"
)
QA_PROMPT = QuestionAnswerPrompt(QA_PROMPT_TMPL)
formatted_prompt = QA_PROMPT.format(context_str=context_str, query_str=query_str)
print(formatted_prompt)context_str and query_str values filled in. The context_str value will be displayed between the dashed lines in the output.Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:context_str is a string containing the context information that you want to provide to the model. In the examples provided in the extracts, the context_str is manually set or retrieved from an index.context_str is set manually:context_str = "The author grew up in a small town, playing sports and reading books."
context_str is set manually for a specific table:city_stats_text = (
"This table gives information regarding the population and country of a given city.\n"
"The user will query with codewords, where 'foo' corresponds to population and 'bar'"
"corresponds to city."
)
table_context_dict = {"city_stats": city_stats_text}context_str can also be retrieved from an index using the query_index_for_context function:query_str = "Which city has the highest population?" context_str = query_index_for_context(table_schema_index, query_str)
context_str is either manually set or retrieved from an index, and then used in the prompt template.Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:context_str is retrieved from an index using the query_index_for_context function. In the SQL guide (https://gpt-index.readthedocs.io/en/latest/guides/tutorials/sql_guide.html), the context_str is retrieved as follows:query_str = "Which city has the highest population?"
context_str = SQLContextContainerBuilder.query_index_for_context(
table_schema_index,
query_str,
store_context_str=True
)context_str to the console, you can simply use the print function:print(context_str)
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:GPTSimpleVectorIndex is a class in the GPT Index library that provides an in-memory vector store for semantic search. It allows you to create an index from a collection of documents and then query the index using natural language questions. The index uses embeddings for the input text chunks and returns relevant responses based on the query.GPTSimpleVectorIndex:from gpt_index import GPTSimpleVectorIndex, SimpleDirectoryReader
# Load documents, build the GPTSimpleVectorIndex
documents = SimpleDirectoryReader('../paul_graham_essay/data').load_data()
index = GPTSimpleVectorIndex.from_documents(documents)
# Query index
response = index.query("What did the author do growing up?")# Save index to disk
index.save_to_disk('index_simple.json')
# Load index from disk
index = GPTSimpleVectorIndex.load_from_disk('index_simple.json')Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:GPTSimpleVectorIndex does not have a context_str that it sets directly. The GPTSimpleVectorIndex is used for semantic search over a collection of documents. It stores document embeddings and retrieves relevant documents based on the query.context_str is used in the context of custom prompts, like the QuestionAnswerPrompt, where you provide context information to the model. The context_str is typically set manually or retrieved from an index, such as in the SQL guide (https://gpt-index.readthedocs.io/en/latest/guides/tutorials/sql_guide.html) example.context_str with GPTSimpleVectorIndex, you would need to manually set the context or retrieve it from another source, and then use it in your custom prompt.Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: