Build AI methods with your coding agent, then run them anywhere: as an MCP for chatbots, as a webapp for people, or via API for your software.
description: "Control whether Pipelex fetches address-referenced method packages from the network when they are missing from the installed methods."
Methods Configuration
The MethodsConfig class controls the behavior of installed and fetched method packages.
Configuration Options
class MethodsConfig(ConfigModel):
fetch_on_miss: bool
Fields
fetch_on_miss: When a bundle references another method by address (github.com/...->domain.pipe) and no installed method matches, fetch the package by address and install it into~/.mthds/methods/so the load can proceed. When disabled, such a miss refuses the bundle with anunresolved_package_dependencyitem naming the address and how to ship the package with the bundle or install it, and the network is never touched at load time.
Example Configuration
[interpreter.methods]
fetch_on_miss = true
Environment Variable Override
The PIPELEX_METHODS_FETCH_ON_MISS environment variable overrides the config when set: 1/true/yes/on enables fetching, 0/false/no/off disables it. An unrecognized value is warned about and ignored in favor of the config.
# Force-disable network fetches at load time for this shell
export PIPELEX_METHODS_FETCH_ON_MISS=0
When to Disable It
- Offline or air-gapped environments, where a fetch could only hang or fail.
- Deployments that must never pull code from the network at load time.
- Reproducibility-sensitive runs where every dependency must be pre-installed and reviewed.
Related Documentation
- Packages — Cross-package references and include by address
- Run a Method by Address — The reference grammar, bounds, and provenance