{
    "languageModel": "AZURE_GPT_4o_2024_1120",
    "companyNameExtraction": {
        "systemMessage": "You are given a file name. Your task is to extract the name of the company contained in the file name. If you can not extract a company name, return \"NONE\". \"IC\" is not a company name.\n\nThe expected output is a JSON in the format: \n{\n    \"company_name\": \"<extracted name of company>\"\n}",
        "userMessage": "$document_name",
        "exampleMessages": [
            {
                "role": "user",
                "content": "20240215_annual_report_ubs.pdf"
            },
            {
                "role": "assistant",
                "content": "{\n    \"company_name\": \"UBS\"\n}"
            },
            {
                "role": "user",
                "content": "FIRST_QUARTER_UNIQUE_2022.pdf"
            },
            {
                "role": "assistant",
                "content": "{\n    \"company_name\": \"Unique\"\n}"
            },
            {
                "role": "user",
                "content": "2022_03_15_Quarterly_Report.pdf"
            },
            {
                "role": "assistant",
                "content": "{\n    \"company_name\": \"NONE\"\n}"
            },
            {
                "role": "user",
                "content": "memo_2024 (vf).pdf"
            },
            {
                "role": "assistant",
                "content": "{\n    \"company_name\": \"NONE\"\n}"
            }
        ]
    },
    "questionExtraction": {
        "useCompanyName": true,
        "additionalInputKeywords": [],
        "dummyExtendedQuestionPlaceholder": "Additional Input",
        "sheet1Name": "Sheet1",
        "sheet2Name": "Sheet2",
        "originalQuestion": {
            "system": "Your role as an assistant is to analyze a transcript from a meeting about an investment opportunity. Your task is to identify and highlight all the questions posed by the participants. This will include both direct questions and requests for more information or clarity. Here's a step-by-step guide to your task:\n\n**Name Extraction:** Identify the name of the individual who asked the question. The name is always the first word in the bullet point. If there is no name mentioned, assign the bullet point name to \"Unknown\".\n\n**Question Extraction:** Identify and extract the question or request for information. Questions are sentences ending with a question mark, or requests for clarification or information (e.g., \"We need to understand why this happened\", \"We should look in this\"). Extract the question or request verbatim, so it can be located in the transcript using a simple text search (ctrl+F).\n\n**Multiple Questions:** If a single person asks several consecutive questions within one bullet point, extract each of them separately. They will be highlighted differently for easy identification.\n\n**Question Mark Rule:** Any sentence ending with a question mark should be considered a question and must be included in the output.\n\nRemember, your primary goal is to identify and extract the questions or requests from the transcript. The extracted text should be as concise as possible, focusing only on the sentences of interest. There's no need for additional context as this will be handled at a later stage.\n\nPlease structure your output as a JSON in the following format:\n\n{\n    \"question\": [\n        {\"name\": \"Alice\",\"original_question\": \"What is photosynthesis?\"},\n        {\"name\": \"Bob\",\"original_question\": \"Who was Cleopatra?\"},\n        {\"name\": \"Bob\",\"original_question\": \"I would like to understand better the market price of Novartis.\"},\n        {\"name\": \"Bob\",\"original_question\": \"We have to look up why this company grew so much in the last 2 years.\"},\n        {\"name\": \"Bob\",\"original_question\": \"You need to review the tech stack and understand how old it really is.\"},\n        {\"name\": \"Bob\",\"original_question\": \"You need to understand why this happened.\"}\n    ]\n}\n\nYou must not create any questions. Only extract the questions directly from the text, word by word, without correcting typos or punctuations. If no questions are present, return an empty question list.\n",
            "systemWithStructuredOutput": "Your role as an assistant is to analyze a transcript from a meeting about an investment opportunity. Your task is to identify and highlight all the questions and action points posed by the participants. This will include both direct questions and requests for more information or clarity. Here's a step-by-step guide to your task:\n\n**Name Extraction:** Identify the name of the individual who asked the question. The name is always the first word in the bullet point. If there is no name mentioned, assign the bullet point name to \"Unknown\". 'Chair' is not a name. Do not assign 'Chair' as a name.\n\n**Transcript Extraction:** Identify and extract transcript sections matching at least one category explained below. Extractsentence verbatim, so it can be located in the transcript using a simple text search (ctrl+F). Be sure not to miss any transcript section. Rather extract a sentence to much than miss one.\nHere a definition of the extract item categories:\n${categories}\n\n**Multiple Questions:** If a single person asks several consecutive questions within one bullet point, extract each of them separately. They will be highlighted differently for easy identification.\n\n**Question Mark Rule:** Any sentence ending with a question mark should be considered a question and must be included in the output.\n\nYou must not create any questions. Only extract the text sections directly from the provided transcript, word by word, without correcting typos or punctuations. If no questions are present, return an empty text extract list. Never create any extractions.\n",
            "trigger": "Transcript:\n'''\n${chunk}\n'''\n\nEvery question or request extracted from the transcript SHOULD be confined to a maximum of TWO sentences (preferably one sentence).\n\nYour answer in JSON format:",
            "triggerWithStructuredOutput": "Transcript:\n'''\n${chunk}\n'''\n\nEvery extract from the transcript SHOULD be confined to a maximum of TWO sentences (preferably one sentence). Again, make sure not to miss any relevant transcript section. Never create any extractions by yourself."
        },
        "additionalInputs": {
            "system": "Your purpose is to convert a text with bullet points into a structured JSON object. Each bullet point represents a contribution from an individual. When it is not clear to whom a bullet point belongs, assign it to the last person mentioned. If there is no name mentioned, assign the bullet point name to \"Unknown\".\n\nStructure the output in JSON, associating each bullet point with the respective contributor\"s name. Each bullet point should be listed as a separate entry. Ensure that you do not break or shorten the bullet points; include each bullet point in its entirety. Maintain the keys \"question\", \"name\", and \"original_question\" in your JSON structure.\n\nJSON Output:\n{\n  \"question\": [\n    {\n      \"name\": \"Alice\",\n      \"original_question\": \"What is photosynthesis?\"\n    },\n    {\n      \"name\": \"Alice\",\n      \"original_question\": \"25 years of investment is a lot\"\n    },\n    {\n      \"name\": \"Bob\",\n      \"original_question\": \"Who was Cleopatra?\"\n    }\n  ]\n}\n\nExtract the bullet points directly from the text, word by word, without correcting typos or punctuations. If no bullet points are present, return an empty question list. Do not break up the bullet points. Include each bullet point in its entirety as a single entry in the \"original_question\" field.\n",
            "trigger": "Input:\n'''\n${chunk}\n'''\n\nJSON output:"
        },
        "extendedQuestion": {
            "system": "You are a helpful AI designed to augment a given question extracted from a transcript, to enhance its clarity and comprehensibility. The original question is provided as well as the context in which it was asked. Your task consist of focusing on improving the question by enriching it with relevant details according to the context. You MUST be concise and you must use as little words as possible. \n\nStructure your output as a JSON in the following format:\n{\n    \"extended_question\": \"A concise augmented question with only relevant details from context. \"\n}\n",
            "triggerWith": "Question:\n'''\n${original_question}\n'''\n\nContext:\n'''\n${chunk}\n'''\n\nThe transcript pertains to the company ${company_name}. If \"they\" or \"it\" is mentioned and the reference is ambiguous, assume that ${company_name} is being referred to.\n\nJSON output:\n",
            "triggerWithout": "Question:\n'''\n${original_question}\n'''\n\nContext:\n'''\n${chunk}\n'''\n\nJSON output:\n",
            "useCompanyName": false
        },
        "topic": {
            "system": "Your task is to give assign a topic to a \"question\" that has been asked during an investment meeting. The topic should be one of the following:\n\n[\n'Artificial Intelligence',\n'Business Plan / Underwriting',\n'Management',\n'Customers',\n'Competitors,\n'Financials',\n'Financing',\n'Go To Market / Sales',\n'Governance',\n'Growth',\n'Integration',\n'Market',\n'M&A (Mergers and Acquisitions)',\n'Product',\n'Strategy',\n'Technology',\n'Valuation'\n]\n\nIf you can not find a topic for the question, put the topic as 'Other'.\n\nTo help you achieve the task, you will be provided the question as well as the transcript where the question was extracted.\n\nStructure your output as a JSON in the following format:\n{\n    \"explanation\": \"A brief explanation of why you chose this topic\",\n    \"topic\": \"topic that describes best the question. Only choose one.\"\n}",
            "trigger": "Question:\n'''\n${original_question}\n'''\n\nContext:\n'''\n${chunk}\n'''\n\nJSON output:\n"
        },
        "keywordSearchMode": "basic",
        "enableStructuredOutput": false,
        "chunkSplitAtLineBreaksForQuestionExtraction": false
    },
    "excelGenerator": {
        "uploadScopeId": null,
        "uploadToChat": true,
        "renameColMap": null,
        "tableHeaderFormat": {
            "bg_color": "#966919",
            "bold": true,
            "font_color": "white",
            "text_wrap": true
        },
        "skipIngestion": true,
        "tableDataFormat": {
            "bg_color": "#FFFFFF",
            "bold": false,
            "font_color": "black",
            "text_wrap": true,
            "border": 1,
            "valign": "top"
        }
    }
}