retool-complex-demand-analyzer

This is a demo repository with a segment of a solution composed of a Retool internal App that helps teams quickly understand complex client demands and speeds up response times—from an average of 80 minutes per demand to 15 minutes.

Objective

Although this project is private, the demo code samples clearly illustrate its simplicity and show how easily the solution can be replicated by anyone with access to Retool and the OpenAI API.

Brief Context

Expert networks depend on rapidly understanding client requests. To accelerate this process, I built an app that integrates:

• Client-submitted forms hosted on a private app.
• Automated context summarization using OpenAI.
• A Retool frontend for quick internal review.

This solution helps teams quickly understand complex client demands, reducing processing time significantly.

Process Flow Diagram

flowchart TB
    A[1 - New client demand<br>received in private app] --> B[2 - Retool uses GraphQL<br>to parse property values]
    B --> C[3 - Parsed data stored<br>in Retool table]
    C --> D[4 - JS script calls OpenAI agent<br>for new form inputs]
    D --> E[5 - Retool sends inputs<br>to agent for suggested response]
    E --> F[6 - OpenAI response<br>saved in Retool DB]
    F --> G[7 - Internal user reviews<br>and can request revision]
    G --> H[8 - Feedback + demand<br>sent back to OpenAI]
    H --> I[9 - Approved data<br>saved to private app]

Loading

Folder Structure

├── README.md
├── graphql/
│   ├── fetchBananaOrders.graphql        # Query to retrieve sanitized banana order data
│   └── updateBananaOrder.graphql          # Mutation to update a banana order in the private app
├── sql/
│   └── updateDemandPostgresDatabase.sql   # SQL query to update banana orders in the Postgres database
├── js/
│   ├── bananaOrderProcessor.js            # Processes fetched data and calls OpenAI for updated details
│   └── orchestrator.js                    # Orchestrates updates between the GraphQL API and Postgres database

Draft Solution

1. When a new client demand is received in our private app, Retool uses GraphQL to parse the relevant property values.
2. The parsed data is stored in a Retool table, referencing our product objects through foreign keys.
3. A JavaScript function then calls an OpenAI agent, requesting inputs for a new internal-client demand form based on predefined models.
4. Retool sends these inputs to the agent, asking for a suggested response.
5. The OpenAI-generated response is stored in the Retool database.
6. An internal user reviews the response and can request a revised version by clicking a button in the Retool frontend.
7. The user may provide feedback or additional input in an open text field, which is sent along with the client demand details.
8. Once approved, the data inputs—updated with OpenAI’s feedback—are saved back to our private app.