Fabric Bridge

Run Fabric AI patterns via the fabric-ai CLI. Each pattern is a reusable system prompt for a specific task.

See references/popular-patterns.md for a curated list of high-quality patterns by category.

Important Notes

• The command is fabric-ai, not fabric.
• First-time setup: run fabric-ai -S to configure API keys.
• If pattern list is empty: run fabric-ai -U to update patterns.
• Use -s (stream) for most calls to avoid long waits.

Core Commands

Basic usage

echo "input text" | fabric-ai -p \x3Cpattern>

Stream output (recommended)

echo "input text" | fabric-ai -p \x3Cpattern> -s

Process a YouTube video

fabric-ai -y "https://youtube.com/watch?v=..." -p extract_wisdom -s

Process a web page

fabric-ai -u "https://example.com/article" -p summarize -s

Specify model

echo "input" | fabric-ai -p \x3Cpattern> -m gpt-4o

Chinese output

echo "input" | fabric-ai -p \x3Cpattern> -g zh -s

Chain patterns (pipe output to next pattern)

echo "input" | fabric-ai -p extract_wisdom | fabric-ai -p summarize

Reasoning strategy (requires setup)

echo "input" | fabric-ai -p \x3Cpattern> --strategy cot -s

Process an image (multimodal)

echo "describe this image" | fabric-ai -p \x3Cpattern> -a /path/to/image.png -s

Use context

echo "input" | fabric-ai -p \x3Cpattern> -C my_context -s

Session continuity

echo "input" | fabric-ai -p \x3Cpattern> --session my_session -s

Save output to file

echo "input" | fabric-ai -p extract_wisdom -o output.md

Copy output to clipboard

echo "input" | fabric-ai -p extract_wisdom -c

Dry run (preview without calling API)

fabric-ai -p \x3Cpattern> --dry-run

List all available patterns

fabric-ai -l

Template Variables

Patterns can contain {{variable}} placeholders. Pass values with -v:

# Single variable
echo "input" | fabric-ai -p \x3Cpattern> -v="#role:expert"

# Multiple variables
echo "input" | fabric-ai -p \x3Cpattern> -v="#role:expert" -v="#points:30"

Custom Patterns

Create custom patterns at ~/.config/fabric/patterns/\x3Cname>/system.md.

Each pattern directory contains a system.md file with the system prompt.

Feeding File Content

cat file.txt | fabric-ai -p \x3Cpattern> -s
cat file1.md file2.md | fabric-ai -p \x3Cpattern> -s

Tips

• Prefer -s (stream) for interactive use — output appears incrementally.
• Chain patterns for multi-step processing (extract → summarize → translate).
• Use -g zh when the user wants Chinese output.
• Use -o file.md to save output, -c to copy to clipboard.
• Use --dry-run to inspect what will be sent before making API calls.
• Run fabric-ai -U periodically to get new community patterns.