Skip to main content

​
Overview

The Pricing Intelligence platform provides a natural language chat interface powered by the H.A.R.V.E.Y. (Holistic Analysis and Regulation Virtual Expert for You) agent. This guide walks you through the basics of interacting with the system.

​
Accessing the Interface

Once you have the system running via Docker Compose, navigate to the frontend: 
http://localhost:5173
You’ll see the H.A.R.V.E.Y. Pricing Assistant chat interface with:
  • A chat transcript area displaying conversation history
  • A control panel for composing queries and managing context
  • Options to upload YAML files or provide pricing URLs

​
Chat Interface Components

1

Question Input

Type your pricing questions in natural language in the text area at the bottom of the screen.
2

Context Management

Add pricing context by:
  • Uploading YAML files (drag-and-drop or file picker)
  • Providing pricing URLs (detected automatically or added manually)
  • Using preset examples from the prompt library
3

Submit Query

Click the submit button or press Enter to send your question to H.A.R.V.E.Y.
4

Review Response

H.A.R.V.E.Y. will analyze the pricing model and provide a grounded answer based on the extracted data.

​
Example Queries

Here are common types of questions you can ask:

​
Basic Plan Information

What plans are available in the Overleaf pricing model?

​
Comparing Plans

Compare the features of the STANDARD and PROFESSIONAL plans.

​
Usage Limits

What is the maximum number of collaborators allowed per project in each plan?

​
Complex Questions

What is the cheapest plan that includes GitHub integration and supports 
more than 5 collaborators per project?

​
Working with URLs

H.A.R.V.E.Y. can automatically extract pricing data from SaaS websites:
1

Paste a URL in your question

Include the pricing page URL directly in your question:
What is the cheapest plan for Buffer (https://buffer.com/pricing) 
that includes 10 channels?
2

Automatic extraction

H.A.R.V.E.Y. detects the URL, calls the iPricing tool to extract the pricing model, and adds it to the conversation context.
3

Query the data

Once extracted, you can ask follow-up questions about the pricing model without repeating the URL.
URL extraction happens automatically when you include a URL in your question. The system uses the A-MINT API to transform the webpage into a structured Pricing2Yaml format.

​
Using Preset Prompts

The interface includes pre-configured example queries to help you get started:
  1. Find best value plan with GitHub integration - Demonstrates optimal plan selection with specific feature requirements
  2. Compare collaborator limits across plans - Shows how to analyze usage limits
  3. Analyze pricing for redundant plans - Illustrates validation and diagnostics
Click on any preset to automatically populate the question field and load the associated YAML context.

​
Uploading YAML Files

If you already have a Pricing2Yaml file, you can upload it directly:
1

Click the upload button

Located in the control panel
2

Select YAML file(s)

You can upload multiple files at once. Each file is validated and added to the conversation context.
3

Verify upload

Uploaded files appear in the context panel with their filename labels.
4

Query the data

Ask questions referencing the uploaded pricing models by name or features.

​
Managing Context

The context panel shows all active pricing models in the current conversation:
  • URL Context (🌐): Pricing extracted from URLs with transformation status
  • YAML Context (📄): Uploaded or preset YAML files
  • Remove Items: Click the × button next to any context item to remove it
  • Clear All: Use the “Clear Context” button to remove all items and start fresh
When you clear context or start a new conversation, all uploaded YAML files and extracted pricing models are removed. You’ll need to re-upload or re-extract them if needed.

​
Starting a New Conversation

Click the “New conversation” button in the header to:
  • Clear all messages from the chat transcript
  • Remove all context items (URLs and YAML files)
  • Reset the question input field
  • Start with a clean slate

​
Theme Toggle

Switch between light and dark mode using the theme toggle button in the header. Your preference is saved to browser localStorage.

​
Response Metadata

H.A.R.V.E.Y.’s responses may include additional metadata:
  • Plan: Shows the reasoning steps H.A.R.V.E.Y. took to answer your question
  • Result: Raw analysis results from the MCP tools (optimal, subscriptions, validate, etc.)
Metadata provides transparency into how H.A.R.V.E.Y. arrived at its answer, showing tool invocations and intermediate results.

​
Best Practices

Be Specific

Include specific feature names, usage limits, or constraints in your questions for more accurate results.

One Model at a Time

For best results, focus on one pricing model per question, or explicitly compare two models.

Use Exact Names

When referencing features or plans, use the exact names from the YAML schema for grounded responses.

Follow-up Questions

Context persists across the conversation, so you can ask follow-up questions without repeating information.

​
Troubleshooting

​
URL Extraction Fails

If H.A.R.V.E.Y. cannot extract pricing from a URL:
  • Verify the URL is accessible and points to a pricing page
  • Check that the page contains structured pricing information
  • The A-MINT service uses OpenAI’s models and may fail on complex or non-standard layouts

​
No Response or Error

If you receive an error message:
  • Check that all backend services are running (docker-compose ps)
  • Verify your OpenAI API keys are configured correctly
  • Review the browser console for network errors
  • Check service logs: docker-compose logs harvey-api

​
Empty or Incomplete Answers

If H.A.R.V.E.Y. provides incomplete information:
  • Rephrase your question to be more specific
  • Verify the YAML file contains the requested data
  • Check that features/limits are correctly named in the schema

​
Next Steps

Pricing Extraction

Learn how to use the iPricing tool to extract structured pricing data from URLs

Optimization Queries

Discover advanced optimization features for finding optimal plans

Build docs developers (and LLMs) love

Get started for freeTalk to us