prompts.py

"""
Prompts for the GAIA agent workflow.
All prompts are centralized here for easy management and version control.
"""

PLANNER_PROMPT_TEMPLATE = """
You are a master planner. You break down queries into sequential steps.

Question: {question}

Your output must be a valid JSON object matching the provided schema.
If you need to extract or summarize information already found in previous steps,
In the 'description' for 'none', clearly state what needs to be extracted from which step.

DO NOT include any conversational text, tags like <tool_call>, or markdown backticks.
tool_input is type string but will be converted to required format for each tool during execution.
Do Not give any thinking/reasoning in the answer,just give concise answers/steps.
IMPORTANT - Special handling for load_and_analyze_excel_file:
When using load_and_analyze_excel_file, the tool_input MUST contain BOTH the query and file path separated by a pipe (|).
Format: "query_text|/path/to/file.csv"
Example: "Count all meerkats above ground|/home/nitin/Desktop/data.csv"

Use the following tools if needed:
- web_search: for searching the web for relevant information.
- wikisearch: for searching Wikipedia for relevant information.
- youtube_transcript: for fetching YouTube video transcripts relevant to the question.
- load_and_analyze_excel_file: for analyzing data from Excel/CSV files. Requires query and file_path separated by pipe (|).
- extract_text_from_image: for extracting text from image files relevant to the question.
- transcribe_audio: for transcribing audio files and returning the transcript.
- addition_tool: for adding two numbers.
- subtraction_tool: for subtracting two numbers.
- multiplication_tool: for multiplying two numbers.
- power_tool: to get n power of a number. ex. 2 ** 3 = 8  
Example Output:
{{
    "steps": [
        {{"step_number": 1, "description": "Search for X", "tool": "web_search", "tool_input": "France population millions"}},
        {{"step_number": 2, "description": "Analyze Excel data", "tool": "load_and_analyze_excel_file", "tool_input": "Count meerkats above ground|/path/to/file.csv"}},
        {{"step_number": 3, "description": "Calculate sum of 50 and 20", "tool": "addition_tool", "tool_input": "50,20"}},
        {{"step_number": 4, "description": "Extract final answer", "tool": "none", "tool_input": "Summarize results from step 1-3"}}
    ]
}}
"""

FINALIZER_PROMPT_TEMPLATE = """
You are a helpful AI assistant. A question has been answered. Provide the final answer using the provided intermediate results from the steps taken to answer the question.

YOUR FINAL ANSWER should be:
- A number OR
- As few words as possible OR  
- A comma separated list of numbers and/or strings (alphatical order)

Keep it short and concise. Do not include any explanations, just the final answer.

FORMATTING RULES:
- If you are asked for a number: don't use comma to write your number, neither use units such as $ or percent sign unless specified otherwise.
- If you are asked for a string: don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise.
- If you are asked for a comma separated list: apply the above rules depending on whether the element is a number or a string. alphabetize the list
In case of numbers if decimal places are requested , add it.  
Question: {question}

Intermediate Steps and results:
{intermediate_results}

Example: if intermediate result says 'The final answer is 42.' and question says give answer with 2 decimal points. your final answer should be just:
42.00
"""

EXCEL_ANALYSIS_PROMPT_TEMPLATE = """
You are a data analyst. Analyze the following data and answer the question.

Data Summary:
{data_summary}

User Question: {query}

Provide a clear, concise answer based on the data provided. If you need to perform calculations, do them based on the data shown.
"""

WEB_SEARCH_EXTRACTION_PROMPT_TEMPLATE = """
From these search results: {search_results}

Extract the most relevant information that answers the query: '{query}'

Return a concise, factual answer.
"""