Skip to main content
Version: v1.6.1 🚧

Create a prompt template

Overview​

A prompt template allows users to guide the language model in generating specific, tailored text for a Collection. For example, a prompt template can be used to customize the description of a Collection.

It is a structured outline with instructions that ensure consistent and relevant outputs. By providing reusable instructions alongside the user's query, a prompt template helps customize responses from h2oGPTe, ensuring the output meets specific needs.

Example​

from h2ogpte import H2OGPTE

client = H2OGPTE(
    address="https://h2ogpte.genai.h2o.ai",
    api_key='sk-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX',
)

prompt_template_id = client.create_prompt_template(
    name="My prompt template",
    description="My first prompt template",
    lang="en",
    system_prompt="You are h2oGPTe, an expert question-answering AI system created by H2O.ai.",
    pre_prompt_query=(
        "Pay attention and remember the information below. You will need to use only any "
        "chat history, any images given, or any document text in order to answer the "
        "question or imperative at the end."
    ),
    prompt_query=(
        """According to only the information in any chat history, any images given, or any document """
        """text provided within the context above, give a well-structured response (that starts with """
        """ "According to") to:"""
    ),
    hyde_no_rag_llm_prompt_extension="Keep the answer brief, and list the 5 most relevant key words at the end.",
    pre_prompt_summary=(
        "In order to write a concise single-paragraph or bulleted list summary, pay attention "
        "to any chat history, any images given, or any following text:"
    ),
    prompt_summary=(
        "Using only any chat history, any images given, or any text above, write a condensed "
        "and concise well-structured Markdown summary of key results."
    ),
    system_prompt_reflection=(
        "You are acting as a judge. You must be fair and impartial and pay attention to details."
    ),
    prompt_reflection=(
        """Review the user's question and the corresponding response using the additive 5-point """
        """scoring system described below. Points are accumulated based on the satisfaction of """
        """each criterion:\n"""
        """- Add 1 point if the response is relevant and provides some information related to the """
        """user's inquiry, even if it is incomplete or contains some irrelevant content.\n"""
        """- Add another point if the response addresses a substantial portion of the user's question, """
        """but does not completely resolve the query or provide a direct answer.\n"""
        """- Award a third point if the response answers the basic elements of the user's question in """
        """a useful way, regardless of whether it seems to have been written by an AI Assistant or """
        """if it has elements typically found in blogs or search results.\n"""
        """- Grant a fourth point if the response is clearly written from an AI Assistant's """
        """perspective, addressing the user's question directly and comprehensively, and is """
        """well-organized and helpful, even if there is slight room for improvement in clarity, """
        """conciseness or focus.\n"""
        """- Bestow a fifth point for a response that is impeccably tailored to the user's question """
        """by an AI Assistant, without extraneous information, reflecting expert knowledge, and """
        """demonstrating a high-quality, engaging, and insightful answer.\n"""
        "\n<user>\n%s\n</user>\n\n<response>\n%s\n</response>\n\n"
        """After examining the user's instruction and the response:\n\n"""
        """- Briefly justify your total score, up to 100 words.\n"""
        """- Conclude with the score using the format: "Score: <total points> / 5"\n\n"""
        """Remember to assess from the AI Assistant perspective, utilizing web search knowledge as """
        """necessary. To evaluate the response in alignment with this additive scoring model, we'll """
        """systematically attribute points based on the outlined criteria."""
    ),
    auto_gen_description_prompt=(
        "Create a short one-sentence summary from the above context, with the goal to make it "
        "clear to the reader what this is about."
    ),
    auto_gen_document_summary_pre_prompt_summary=(
        "In order to write a concise single-paragraph summary, pay attention to the following text:"
    ),
    auto_gen_document_summary_prompt_summary=(
        "Using only the text above, write a condensed and concise summary of key results "
        "(preferably as one paragraph)."
    ),
    auto_gen_document_sample_questions_prompt=(
        'Create 10 interesting questions for which the answers are contained in the information '
        'provided above. Respond in JSON, like this: {{"q1": "Question 1?", "q2": "Question 2?", ..., "q10": "Question 10?"}}.'
    ),
    default_sample_questions=[
        "Create five good questions",
        "Summarize in one sentence",
        "Explain this to a 5-year old",
        "Tell me something interesting",
    ],
    image_batch_image_prompt=(
        """<response_instructions>\n"""
        """- Act as a keen observer with a sharp eye for detail.\n"""
        """- Analyze the content within the images.\n"""
        """- Provide insights based on your observations.\n"""
        """- Avoid making up facts.\n"""
        """- Do not forget to follow the system prompt.\n"""
        """</response_instructions>"""
    ),
    image_batch_final_prompt=(
        """<response_instructions>\n"""
        """- Check if the answers already given in <image> XML tags are useful.\n"""
        """- Image answers came from a vision model capable of reading text and images within """
        """the images.\n"""
        """- If image answers are useful, preserve all details the image answers provide and use """
        """them to construct a well-structured answer.\n"""
        """- Ignore image answers that had no useful content, because any single batch of images """
        """may not be relevant. Focus on all details from image answers that are relevant and useful.\n"""
        """- Check if the document text can answer the question.\n"""
        """- Check if the chat history can answer the question.\n"""
        """- Check if any figure captions can answer the question.\n"""
        """- If answers conflict between text, chat history, and figure captions, do not focus """
        """your response on this conflict.\n"""
        """- In handling conflicting answers, use logical reasoning and supporting evidence to """
        """assess the plausibility of each answer.\n"""
        """- In handling conflicting answers, choose the most consistent answer -- i.e., the """
        """most common answer among conflicts (self-consistency reasoning) or one that aligns """
        """with well-established facts.\n"""
        """- In handling conflicting answers, one may choose one data source over another -- i.e., """
        """text is probably more reliable than an image when the question can be answered from """
        """text, while an image is more reliable than text for flowcharts, photos, etc.\n"""
        """- Do not forget to follow the system prompt.\n"""
        """- Finally, according to our chat history, the above documents, figure captions, or given """
        """images, construct a well-structured response.\n"""
        """</response_instructions>"""
    ),
)

print(prompt_template_id)
b7f06049-2aae-4db3-84fa-77....
Feedback