Design a meta-prompt that generates detailed instructions for building a LangChain workflow.  This workflow should address a specific business problem, such as automating customer support responses or generating creative marketing content. The meta-prompt should guide the user through the following steps:</p>
<p>1. **Problem Definition:** Clearly define the business problem the LangChain workflow aims to solve.  Include specific metrics for success.<br />
2. **Data Source Identification:** Specify the data sources required (e.g., databases, APIs, files) and their formats.<br />
3. **Chain Design:** Outline the sequence of LangChain modules (LLMs, prompts, chains, indexes) needed, detailing the function of each component and the data flow between them.<br />
4. **Prompt Engineering:** Craft effective prompts for each LLM, considering prompt engineering best practices (e.g., few-shot learning, prompt chaining).<br />
5. **Error Handling:** Include strategies for handling potential errors and unexpected inputs.<br />
6. **Testing and Evaluation:** Define a plan for testing the workflow and evaluating its performance against the predefined success metrics.<br />
7. **Deployment and Monitoring:** Outline how the workflow will be deployed (e.g., locally, in the cloud) and monitored for performance and errors.</p>
<p>The generated instructions should be detailed enough for a developer with intermediate LangChain experience to implement the workflow.  The output should include code snippets (in Python) where appropriate, illustrating the implementation of key components. The meta-prompt should also consider the scalability and maintainability of the resulting workflow.  Provide an example of a LangChain workflow designed to summarize customer feedback from a database and categorize it into predefined themes.</p>
<p>The final output should be a structured document outlining the complete LangChain workflow, including code examples, error handling strategies, and deployment considerations.