Tentacles Manager
octobot_tentacles_manager is responsible for the full lifecycle of OctoBot tentacles — the plugin system that extends OctoBot with trading modes, evaluators, exchange connectors, services, and more. It handles everything from downloading and installing a tentacle bundle to generating the Python import infrastructure that makes tentacles loadable at runtime.
What the manager does
Installation works by downloading a ZIP archive (from a URL or a local path), extracting it, and copying each tentacle into the bot's tentacles/ tree while resolving any cross-tentacle requirements declared in metadata.json. An update pass skips tentacles whose installed version is already current, making it safe to re-run against the same source. Uninstall removes the relevant directories and regenerates the __init__.py import files so the rest of the tree stays consistent.
The generated __init__.py files are a first-class output of this package, not a side effect. Each one is built around a call to check_tentacle_version() that gates the import on the minimum compatible version for the tentacle's origin package. If a tentacle's declared version is too old, its import is silently skipped rather than raising an exception, which means a single outdated tentacle cannot break startup for the others.
A repair operation regenerates missing __init__.py files and folder structure without touching tentacle configs. This is the recovery path for a broken installation where the code is intact but the import machinery has gone out of sync.
Configuration management
The manager reads and writes two distinct kinds of configuration. The first is tentacles_setup_config.json at the profile root, which records which tentacles are activated in a given profile. Evaluators and trading modes are deactivated by default and must be explicitly enabled; services and utility tentacles activate automatically on install. The second is per-tentacle JSON config files, each stored inside the tentacle's own config/ directory as the reference default. When a user customizes a tentacle, a profile-specific copy is written to the profile's specific_config/ folder and takes precedence over the reference at runtime.
Discovery and loading
At startup, OctoBot calls the manager to scan the tentacles/ tree up to three folder levels deep, looking for directories that contain a metadata.json. Each discovered module is parsed into a Tentacle model that tracks the type path, class names, version, origin package, and optional tentacle group. The result is cached in a module-level dict keyed by class name, and everything downstream — activation checks, configuration resolution, documentation loading, resource path lookups — reads from that cache.
Export and distribution
The manager can also produce redistributable artifacts. A pack operation copies or zips a set of tentacles from an installed tree into a bundle, with optional Cython compilation for distributing compiled-only packages. The upload path pushes those artifacts to S3 or Nexus artifact repositories.
CLI
The manager ships with a standalone command-line interface and also exposes a register_tentacles_manager_arguments() function that OctoBot uses to attach tentacle sub-commands to its own argument parser. This lets the same install, update, repair, and pack operations be driven from either the manager's own CLI or from within octobot --install, depending on context.