Build-PromptFromFiles

Combine files from a directory into a single prompt, ready for use with large language models (LLMs)

For background on this project see Building files-to-prompt entirely using Claude 3 Opus.

Installation

Install-Module BuildPromptFromFiles

Usage

Use the Build-PromptFromFiles command to combine all files in one or more directories into a single prompt.

Build-PromptFromFiles path/to/file_or_directory [path/to/another/file_or_directory ...]

This will output the contents of every file in the specified directories, combined into an XML format.

Options

• -RAW - Output the prompt as a single string, with each file preceded by its relative path and separated by ---
• -ignore <patterb> - Specify a pattern to ignore files by. For example -ignore *.txt will ignore all .txt files.

  Build-PromptFromFiles path/to/file_or_directory -ignore *.txt, *.dll

Usage - Build Prompt From GitHub Repo

Use the Build-PromptFromGitHubRepo command to download a GitHub repository and build a prompt from its contents.

Build-PromptFromGitHubRepo dfinke/psai | clip

clip is optional and will copy the prompt to the clipboard. You can then paste it into your favorite LLM. Your token limit may be exceeded if the prompt is too large.

Example

Suppose you have a directory structure like this:

my_directory/
├── file1.txt
├── file2.txt
├── temp.log
└── subdirectory/
    └── file3.txt

Running Build-PromptFromFiles my_directory -RAW will output:

my_directory/file1.txt
---
Contents of file1.txt
---
my_directory/file2.txt
---
Contents of file2.txt
---
my_directory/temp.log
---
Contents of temp.log
---
my_directory/subdirectory/file3.txt
---
Contents of file3.txt
---

If you run Build-PromptFromFiles my_directory -RAW --ignore "*.log", the output will exclude temp.log:` the output will be:

my_directory/file1.txt
---
Contents of file1.txt
---
my_directory/file2.txt
---
Contents of file2.txt
---
my_directory/subdirectory/file3.txt
---
Contents of file3.txt
---

XML Output

Anthropic has provided specific guidelines for optimally structuring prompts to take advantage of Claude's extended context window.

The XML output is generated by Build-PromptFromFiles as the default and follows these guidelines. Here is an example of the XML output:

<documents>
<document index="1">
<source>my_directory/file1.txt</source>
<document_content>
Contents of file1.txt
</document_content>
</document>
<document index="2">
<source>my_directory/file2.txt</source>
<document_content>
Contents of file2.txt
</document_content>
</document>
</documents>