<tool id="interactive_tool_chat_analysis" tool_type="interactive" name="Chat with your data" version="0.2">
    <requirements>
        <container type="docker">
            quay.io/goeckslab/chatanalysis:smolagents
        </container>
    </requirements>
    <entry_points>
        <entry_point name="Chat with $input_dataset.element_identifier" requires_domain="True">
            <port>
                8501
            </port>
            <url>
                /
            </url>
        </entry_point>
    </entry_points>
    <command>
        <![CDATA[
        echo "Starting Streamlit" &&
        echo "working dir now: `pwd`" &&
        mkdir -p generate_files &&
        echo "input dataset: $input_dataset" &&
        echo "input dataset datatype: $dataset_type" &&
        ln -sf $__tool_directory__/.env /ChatAnalysis/.env &&
        streamlit run /ChatAnalysis/smolagents_da.py
            "$__user_email__"
            "$openai_api_key_file"
            "$groq_api_key_file"
            "$chat_history"
            "./generate_files"
            "$input_dataset"
            "$dataset_type"
        ]]>
    </command>
    <configfiles>
        <configfile name="openai_api_key_file"><![CDATA[
            $__user__.extra_preferences.get('chatgpt|api_key')
        ]]></configfile>
        <configfile name="groq_api_key_file"><![CDATA[
            $__user__.extra_preferences.get('groq|api_key')
            ]]>
        </configfile>
    </configfiles>
    <stdio>
        <exit_code range="0:0" level="log"/>
    </stdio>
    <inputs>
        <param name="input_dataset"  type="data" format="csv,tsv,h5ad,xlsx,xls,json,parquet,h5,fa,fasta,vcf,gtf,gff,bed" label="Data you want to chat with" optional="false"/>
        <param name="dataset_type" type="select" label="The type of your dataset" >
            <option value="csv">csv</option>
            <option value="tsv">tsv</option>
            <option value="h5ad">h5ad</option>
            <option value="xlsx">xlsx</option>
            <option value="xls">xls</option>
            <option value="json">json</option>
            <option value="parquet">parquet</option>
            <option value="h5">h5</option>
            <option value="fa">fa</option>
            <option value="fasta">fasta</option>
            <option value="vcf">vcf</option>
            <option value="gtf">gtf</option>
            <option value="gff">gff</option>
            <option value="bed">bed</option>
        </param>  
    </inputs>
    <outputs>
        <data name="chat_history" format="json" label="${tool.name} -- chat history json on ${on_string} -- use to restore chat history next time"/>
        <collection name="generate_files" type="list" label="${tool.name} -- generated files">
            <discover_datasets pattern="__name_and_ext__" directory="generate_files" />
        </collection>
    </outputs>
    <tests>
        <test>
        </test>
    </tests>
    <help>
        Analyze and visualize a dataset with a Chat Agent.

        Start a ChatAnalysis instance after selecting a dataset — we provide free access to some LLMs.

        You are encouraged to use your own API keys, as the provided free access has limits. You can obtain a **free API key** for BambooLLM by signing up at https://pandas-ai.com and a **free API key** for Groq by signing up at https://groq.com.

        If you wish to set the OpenAI API key beforehand, go to **User → Preferences → Manage Information → ChatGPT → OpenAI API Key**. This way, you won’t need to set the OpenAI API key each time you use the app.

        Similarly, to set the Groq API key beforehand, navigate to **User → Preferences → Manage Information → Groq → API Key**.

        Alternatively, you can set the API key directly in the ChatAnalysis instance after it has started.
    </help>
</tool>