Select the Right Tool

Why This Matters

In V7 Go, tools power the logic behind each property.

Once you’ve defined what field you want to extract (the property), the next question is:

How should that field be generated?

The tool you assign determines whether a field is extracted using AI, calculated with code, manually entered, or pulled from an external source.

Core Tools in V7 Go

Here’s an overview of the main tools and how to decide when to use each:

Tool	Best For	Notes
AI Model	Natural language extraction, classification, summarization	Most common tool; pairs with prompts; supports grounding/citations
Python	Conditional logic, calculations, parsing JSON, custom formatting	Great for downstream logic and field validation
Web Search	External data lookup, document references	Rarely used in internal workflows; useful for grounding externally
Manual Input	File uploads, user-selected values, ad-hoc text inputs	Used at the start of workflows or for human-in-the-loop scenarios

How Tools Are Applied

1. AI Model

Used to extract text from leases, classify property types, summarize documents.
Paired with Text, Select, and JSON properties.

2. Python

Used to:

Split text from OCR for large documents
Parse outputs from a JSON property
Conditionally route documents (e.g., by OCR length)
- Typically used with Text, Number, and JSON downstream

3. Manual Input

Used to trigger bundles (e.g., a bid comparison with multiple PDFs).

Also supports user dropdowns or comments at workflow start.

Commonly paired with Select, File Upload, or Text Input

4. Web Search

Used to search external sources on the public internet. Can enrich your data, perform analysis, and extract any public facing data from specific and unspecific URLs.

Paired with Text properties most often for maximum customization.

Choosing the Right Tool

Use this decision tree:

Is the field extracted from natural language?

→ Use AI Model
Does the output depend on logic or formatting?

→ Use Python
Does the user need to manually enter/select it?

→ Use Manual Input
Do you need external context (e.g., public data)?

→ Use Web Search

Combining Tools for Advanced Workflows

“AI into JSON → Python” is a common pattern.

Example:

AI extracts a JSON blob containing Start Date and End Date
Python parses and splits that blob into two Number fields
Result: More accurate extraction with fewer hallucinations

Real-world Use Cases → Selecting the Right Tools

Key Takeaway

Every property needs a tool — choose the one that best fits your logic, data type, and output expectations.

Well-structured agents often use a mix of tools to balance reasoning, cost, and control.