Given a user query and a list of sources, write a response that cites individual
sources as comprehensively as possible.
範例 2
工作說明,瞭解使用者並著重於實用性:
You are an enterprise LLM summarization tool. Your task is to understand the
true intent of a user question in the context of enterprise search and
summarization, and provide a helpful answer to the user's question.
範例 3
工作說明:摘要客戶與助理之間的對話:
Given the conversation between a customer and a helpful assistant with some
search results, create a final answer for the assistant.
額外操作說明範例和提示
額外指示應涵蓋特定重要規定。
下表列出一些您可能會在工作說明後提供的額外指示、每個範例解決的問題類型,以及前言解決問題的原因。
要解決的問題
解決方案
範例
答案需要更符合業務需求
提供額外背景資訊和指示,確保摘要內容符合特定用途和目標對象。
範例 1
Utilize the specific context of the workspace (e.g. meeting notes, public
guidance, FAQ) to provide more accurate and relevant summaries.
範例 2
Summarize customer feedback, focusing on their pain points, feature
request and overall satisfaction. Highlight any actionable insights that
can help improve our product or service.
範例 3
For input documents of troubleshooting website, please summary the
problem statement, step-by-step solutions and any relevant tips or
warnings.
範例 4
"XYZ" is an internal forum for engineers to discuss technical problems,
you can use it to summarize technical issues, proposed solutions and any
unresolved challenges or next steps identified in the discussion.
答案必須符合特定風格
清楚指定風格或語調,以及目標對象。
範例 1
Summarize troubleshooting guide for customer support agent in a clear and
concise manner. The summary should be easy for a non-technical user to
understand.
範例 2
Summarize the technical documents for engineers. Focus on the core
functionality, system architecture, and potential challenges.
答案必須採用特定格式
指定輸出格式
範例 1
Use bullet points for steps, numbered lists for rankings, tables for
comparisons, code block for coding example
範例 2
Summarize the key takeaways in a numbered lists
答案必須簡短
明確指示 LLM 生成「簡潔」或「簡短」的摘要。
您也可以視需要指定字數或句數。
範例 1
Please keep summaries concise and focused, providing only the most
essential information to address the user's query.
範例 2
The answer should be less than 200 words.
答案需要更全面
請 LLM 納入重要細節和重點。
範例
Please ensure key details are included.
包含禁止的主題
定義模型在特定情況下的回應方式。
範例
For political questions, the most helpful way is to politely refuse to
answer the question.
減少幻覺 (不正確的資訊)
強調準確度的重要性,並指示 LLM 嚴格遵守文字中呈現的資訊。
範例 1
Keep the summary accurate, ensuring all claims are verifiable within the given context.
範例 2
Use exact words from the context if possible.
完整前言範例
以下提供更多完整前言的範例,這些前言由工作說明和額外指示組成。
範例 1
要求提供簡潔、準確且相關的摘要,並以易於閱讀的格式呈現。
You are an enterprise LLM summarization tool. Your task is to understand the
true intent of a user question in the context of enterprise search and
summarization, and provide a helpful answer to the user's question. Please keep
summaries concise and focused, providing only the most essential information to
address the user's query.
Please also structure and format the summary by
1) prioritize most relevant and accurate information to user's question
2) highlight critical information
3) structure the response and adapt the formatting to be user friendly (e.g.,
use bullet points for steps, numbered lists for rankings, tables for
comparisons, code block for coding example, etc).
範例 2
根據對話內容,針對顧客的查詢提供簡潔、友善且實用的最終回覆。
Given the conversation between a customer and a helpful assistant with some
search results, create a final answer for the assistant.
The answer should addresses the query accurately and concisely (less than 10
sentences), while also being friendly and helpful. If the search results don't
provide enough information to fully answer the question, suggest additional
resources or steps the customer can take.
範例 3
提供完整且易於理解的答案,並引用所提供的來源。
禮貌地拒絕回答任何政治問題。
Given a user query and a list of sources, write a response that cites individual
sources as comprehensively as possible.
The response should be suitable for a non-expert audience.
For political questions, the response should be a polite refusal to answer the
question.
[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["難以理解","hardToUnderstand","thumb-down"],["資訊或程式碼範例有誤","incorrectInformationOrSampleCode","thumb-down"],["缺少我需要的資訊/範例","missingTheInformationSamplesINeed","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間:2025-09-05 (世界標準時間)。"],[[["\u003cp\u003eCustom preambles allow you to tailor the LLM's behavior, improving the quality of generated answers beyond the default settings.\u003c/p\u003e\n"],["\u003cp\u003eA preamble should include a task description that defines the LLM's role and additional instructions that guide its response.\u003c/p\u003e\n"],["\u003cp\u003ePreamble instructions can specify constraints on word usage, preferred or prohibited topics, as well as the desired style, tone, and format of the response.\u003c/p\u003e\n"],["\u003cp\u003eYou can resolve problems like having the answers be more tailored, shorter, more comprehensive, and in a certain format by applying appropriate preamble instructions.\u003c/p\u003e\n"],["\u003cp\u003eThe process of developing a preamble is iterative and should be refined based on user feedback and model updates, including thorough evaluations across use cases.\u003c/p\u003e\n"]]],[],null,["# About custom preambles\n\nThis page discusses custom preambles and how you can write preambles to improve\nthe quality of generated answers.\n\nThe preamble sets the initial context and expectations for the LLM before\nit processes your input document. The preamble influences the\nquality of the generated summaries. There is a default preamble supplied\nwhenever you call the [answer](/generative-ai-app-builder/docs/reference/rest/v1/projects.locations.dataStores.servingConfigs/answer) method. However, you have the option\nto specify your own preamble instead of using the default.\n\nFor instructions on how to specify the preamble in the answer method call, see\n[Specify a custom preamble](/generative-ai-app-builder/docs/answer#preamble-rest).\n\nFor example, you can use the preamble to do the following:\n\n- Specify words that the model can and can't use.\n\n- Specify topics to focus on or avoid.\n\n- Specify the style, tone, and format of the response.\n\nTailoring the preamble can significantly improve the quality of\nsummaries.\n\nThe preamble should have two parts:\n\n- **The task description** that describes the task you are asking the LLM to\n perform. See [Examples of task descriptions](#task-descriptions).\n\n- **Additional instructions** that the LLM should follow. See [Examples and\n tips for additional instructions](#additional-instructions).\n\nExamples of task descriptions\n-----------------------------\n\nHere are some examples of task descriptions. The scenario is that your\nemployees want answers from a data store that contains many company documents.\n\n### Example 1\n\nTask description to comprehensively cite sources: \n\n Given a user query and a list of sources, write a response that cites individual\n sources as comprehensively as possible.\n\n### Example 2\n\nTask description to understand the user and focus on helpfulness: \n\n You are an enterprise LLM summarization tool. Your task is to understand the\n true intent of a user question in the context of enterprise search and\n summarization, and provide a helpful answer to the user's question.\n\n### Example 3\n\nTask description to summarize a conversation between a customer and an\nassistant: \n\n Given the conversation between a customer and a helpful assistant with some\n search results, create a final answer for the assistant.\n\nExamples and tips for additional instructions\n---------------------------------------------\n\nThe additional instructions should capture your specific key requirements.\n\nThe following table gives examples of additional instructions that you might\nprovide after the task description, what kind of problems each example\naddresses, and why the preamble solves the problem.\n\nExamples of complete preambles\n------------------------------\n\nHere are some more examples of complete preambles, made up of the task\ndescription and the additional instructions.\n\n### Example 1\n\nRequest a concise, accurate, and relevant summary and present it in a\nuser-friendly format. \n\n You are an enterprise LLM summarization tool. Your task is to understand the\n true intent of a user question in the context of enterprise search and\n summarization, and provide a helpful answer to the user's question. Please keep\n summaries concise and focused, providing only the most essential information to\n address the user's query.\n\n Please also structure and format the summary by\n\n 1) prioritize most relevant and accurate information to user's question\n\n 2) highlight critical information\n\n 3) structure the response and adapt the formatting to be user friendly (e.g.,\n use bullet points for steps, numbered lists for rankings, tables for\n comparisons, code block for coding example, etc).\n\n### Example 2\n\nProvide a concise, friendly, and helpful final answer to a customer's query\nbased on a conversation. \n\n Given the conversation between a customer and a helpful assistant with some\n search results, create a final answer for the assistant.\n\n The answer should addresses the query accurately and concisely (less than 10\n sentences), while also being friendly and helpful. If the search results don't\n provide enough information to fully answer the question, suggest additional\n resources or steps the customer can take.\n\n### Example 3\n\nProvide comprehensive and understandable answers and cite given sources.\nPolitely decline to answer any political questions. \n\n Given a user query and a list of sources, write a response that cites individual\n sources as comprehensively as possible.\n\n The response should be suitable for a non-expert audience.\n\n For political questions, the response should be a polite refusal to answer the\n question.\n\nBest practices\n--------------\n\nThe following are some best practices for writing and tuning preambles:\n\n- **Iterative Refinement:** Experiment with different preamble\n variations and observe the impact on the answer quality.\n\n- **User Feedback:** Gather feedback from users to identify recurring issues\n and areas for improvement.\n\n- **Stay Updated:**The effectiveness of preamble tuning can vary\n depending on the model version and the nature of your documents. Continuously\n experiment and refine your approach to achieve optimal results.\n\n- **Thorough Evaluation:** Verifying modified preambles across all intended use\n cases helps identify and mitigate potential biases or unexpected behavior\n that may negatively impact summary quality in certain scenarios."]]