Explain KPI report with multimodal LLMs

<p><strong>KPIs report explanation using Gen AI</strong></p><p>This workflow uses multimodal local LLMs (LLaMa3.2:1b and LLaMa3.2-vision) to explain a KPI report containing information (in the form of different plots and charts) regarding a simulated software company.</p><p></p><p><strong>IMPORTANT</strong>: These local models are significantly simpler compared to other available models (like OpenAI's), hence also their performance are not on the same level. However, the fact that they can be run locally ensures <strong>protection of the company data in every processing step</strong>.</p><p></p><p>Following the instructions below, you can learn how to use the workflow and how to modify it if you need to adapt this workflow to other use cases.</p><p><strong>HOW TO USE THE WORKFLOW:</strong></p><ol><li><p>Run "ollama run <strong>llama3.2:1b" </strong>on your command prompt line to enable the text parsing from a PDF report in Step 1</p></li><li><p>Run "ollama run <strong>llama3.2-vision</strong>"<strong> </strong>always on your command prompt line to enable the plot and chart interpretation in Step 2</p></li><li><p>As suggested, in Step 3 refresh the session before the refinement prompt</p></li><li><p>Export the <strong>report </strong>with the KPIs interpretations and recommendations</p></li></ol><p><strong>Notes if you need a different use case:</strong></p><p>If you need to change the workflow for a different use case, the following recommendations may help:</p><ul><li><p>Modify the text extraction (both cleaning steps and prompt) to fit the schema of the new document</p></li><li><p>Explore different models to optimize the performance in the new task</p></li><li><p>Adapt the following prompts for the explanation</p></li><li><p>Modify the output report as needed to fit your use case</p></li></ul>