llm-templates-fsspec

Load LLM templates from local paths and cloud storage (S3, GCS, Azure) using any supported fsspec URL scheme.

Lear more about Filesystem Spec (fsspec) and it's features in https://filesystem-spec.readthedocs.io/

Installation

Install this plugin in the same environment as LLM.

llm install llm-templates-fsspec

or add it along your project if using a environment manager like poetry or uv

# with Poetry
poetry add llm llm-templates-fsspec
# with UV
uv add llm llm-templates-fsspec

Usage

Load templates from local filesystem

llm -t fsspec:file:///home/user/templates/summary.yaml

Load templates from AWS S3

# Configure AWS credentials first
export AWS_ACCESS_KEY_ID=your-key
export AWS_SECRET_ACCESS_KEY=your-secret

llm -t fsspec:s3://my-bucket/templates/analyze.yaml

Load templates from Google Cloud Storage

# Configure GCP credentials
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json

llm -t fsspec:gs://my-bucket/templates/review.yaml

Load templates from Azure Blob Storage

# Configure Azure credentials
export AZURE_STORAGE_ACCOUNT_NAME=myaccount
export AZURE_STORAGE_ACCOUNT_KEY=mykey

llm -t fsspec:az://my-container/templates/report.yaml

Load templates from HTTP/HTTPS

llm -t fsspec:https://example.com/templates/chatbot.yaml

Load templates from FTP

llm -t fsspec:ftp://ftp.example.com/templates/assistant.yaml

Why fsspec?

The llm-templates-fsspec plugin leverages the robust fsspec library to provide:

• Universal Access: Single interface for templates stored anywhere - local files, S3, GCS, Azure, FTP, HTTP, and more built-in implementations
• Production Ready: Battle-tested library used by pandas, dask, and other major data tools
• Credential Management: Integrates with standard credential chains (AWS CLI, gcloud, environment variables)
• Caching: Built-in caching support for remote templates to improve performance
• Extensibility: Support for new storage backends

Examples

Using templates from local

# Assuming your on this repository cloned in your computer.
cat README.md | llm -t "fsspec:file://$(pwd)/demo/template-demo.yaml"

Using templates from a private S3 bucket

# Template in private bucket with AWS CLI credentials configured
curl https://example.com/data.json | \
  llm -t fsspec:s3://private-bucket/templates/analyze-json.yaml

Caching remote templates locally

# Set cache directory for better performance with remote templates
export LLM_TEMPLATES_FSSPEC_CACHE_DIR=/tmp/llm-template-cache

# First run downloads and caches the template
llm -t fsspec:https://raw.githubusercontent.com/org/repo/main/template.yaml "Analyze this"

# Subsequent runs use the cached versio
llm -t fsspec:https://raw.githubusercontent.com/org/repo/main/template.yaml "Analyze that"

Using templates from a corporate GCS bucket

# With service account key file
export GOOGLE_APPLICATION_CREDENTIALS=/path/to/key.json
llm -t fsspec:gs://company-ai-assets/templates/legal-review.yaml < contract.txt

Loading templates with custom storage options

# Pass additional storage options via environment variable
export LLM_TEMPLATES_FSSPEC_STORAGE_OPTIONS='{"client_kwargs": {"region_name": "eu-west-1"}}'
llm -t fsspec:s3://eu-bucket/templates/gdpr-check.yaml

Configuration

Environment Variables

• LLM_TEMPLATES_FSSPEC_CACHE_DIR: Directory for caching remote templates (optional)
• LLM_TEMPLATES_FSSPEC_STORAGE_OPTIONS: JSON string of additional fsspec storage options (optional)

Credential Configuration

This plugin uses standard credential chains for each cloud provider:

• AWS S3: AWS CLI configuration, environment variables, or IAM roles
• Google Cloud Storage: Application default credentials or service account key file
• Azure Blob Storage: Environment variables or Azure CLI credentials

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment using UV

cd llm-templates-fsspec
uv venv
uv sync
source venv/bin/activate

Now install the dependencies and test dependencies:

llm install -e '.[test]'

To run the tests:

pytest

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

Apache 2.0