Search Files Like You Talk: Introducing the FileFortress AI Command

8 min read AI Search CLI

You already know what you want to find. Now you can ask for it directly.

The new filefortress ai command translates natural language into structured search filters, then runs the same search engine you already trust.

What You Gain
  • Faster first search with less flag memorization
  • Interpreted filters you can inspect before execution
  • A safer workflow with --dry-run and --explain

Table of Contents

How the AI Command Works

Under the hood, the command does three steps:

  1. Builds a schema-aware prompt using your current remote context.
  2. Asks the configured AI provider to return structured JSON filters.
  3. Maps those filters into the normal FileFortress search command path.
# Natural language input
filefortress ai "find images smaller than 5kb"

# Interpreted filters (example)
mediaType = images
sizeMaxBytes = 5120

Real Examples That Work Well

Prompt quality matters. Good prompts include explicit filter intent (media type, size, date, remote name/type).

filefortress ai "find images"
filefortress ai "find images smaller than 5kb"
filefortress ai "find videos modified in last 30 days"
filefortress ai "find documents on onedrive larger than 10mb"

The Safe Workflow: Dry-Run First

Use this pattern when trying a new prompt:

# 1) Validate interpretation
filefortress ai "find images smaller than 5kb" --dry-run

# 2) Add explanation if needed
filefortress ai "find images smaller than 5kb" --dry-run --explain

# 3) Execute
filefortress ai "find images smaller than 5kb"
Why This Workflow Matters

Dry-run mode gives you confidence before running a search at scale and helps you quickly refine prompt wording.

Current Limits and When to Use Other Commands

The AI command is optimized for file-search prompts. It is not a general account assistant.

  • Use ai for filter-oriented search intent.
  • Use remotes list for remote inventory and counts.
  • Use tools configure foundry for provider/model setup.
Practical Tip

If you see provider 400 errors, verify the exact model ID for your endpoint. Generic names may not be valid on all OpenAI-compatible runtimes.

Where to Go Next

Try Your First AI Search

Start with --dry-run --explain, verify interpretation, then execute with confidence.

Open Command Docs Read Query Patterns