Right. But, what is a good way to format that table so that gpt=3.5-turbo-16k will recognize it as a list of documents to use as context?
This is sort of what I am sending (as json) to the model now:
Array (
[0] => Array (
[role] => 'system',
[content] => 'You are a helpful assistant. Your goal is to answer each question, using the associated texts as context, as truthfully as you can. If you cannot answer the question or find the meaning or locate the documents requested, tell the user to try re-phrasing the question. Also tell the user to consider a keyword search. In addition, remind user to keep in mind that the the user access may not allow the user to view the requisite documents.'
),
[1] => Array (
[role] => 'user',
[content] => 'Question: Are there any documents that mention Mark\'s IP address?
Number of context documents: 2
Context document 1:
Document Title: \'Apache Server Logs not showing Actual IP Addresses\'
Content: <The detailed content here...>
Source document summary:
Event Date:
Document Groups: D8 Development, D7 Development, Documentation, Development, D9 Development, Site Creation, D10 Development, AI, Real Estate
Document Taxonomy/Tags:
URL: https://master1and1-9.oursite.com/node/911
Questions that this document answers: -----
Context document 2:
Document Title: \'Mark\'s New IP Address\'
Content: <The detailed content here...>
Source document summary:
Event Date:
Document Groups: Development, D9 Development
Document Taxonomy/Tags:
URL: https://master1and1-9.oursite.com/node/1018
Questions that this document answers: -----'
)
)
Do you think it would recognize it better if I restructured like this?
Array (
[0] => Array (
[role] => 'system',
[content] => Array (
[instructions] => 'You are a helpful assistant. Your goal is to answer each question, using the associated texts as context, as truthfully as you can. If you cannot answer the question or find the meaning or locate the documents requested, tell the user to try re-phrasing the question. Also tell the user to consider a keyword search. In addition, remind user to keep in mind that the the user access may not allow the user to view the requisite documents.'
)
),
[1] => Array (
[role] => 'user',
[content] => Array (
[questions] => Array (
[0] => Array (
[text] => 'Are there any documents that mention Mark\'s IP address?',
[context_document_count] => 2
)
),
[context_documents] => Array (
[0] => Array (
[title] => 'Apache Server Logs not showing Actual IP Addresses',
[content] => '<The detailed content here...>',
[summary] => '',
[event_date] => '',
[document_groups] => Array('D8 Development', 'D7 Development', 'Documentation', 'Development', 'D9 Development', 'Site Creation', 'D10 Development', 'AI', 'Real Estate'),
[taxonomy_tags] => Array(),
[url] => 'https://master1and1-9.oursite.com/node/911',
[answers_questions] => Array()
),
[1] => Array (
[title] => 'Mark\'s New IP Address',
[content] => '<The detailed content here...>',
[summary] => '',
[event_date] => '',
[document_groups] => Array('Development', 'D9 Development'),
[taxonomy_tags] => Array(),
[url] => 'https://master1and1-9.oursite.com/node/1018',
[answers_questions] => Array()
)
)
)
)
)
Has anyone tried sending a prompt like this, successfully? Can gpt-3.5-turbo-16k read it like this?