corerun

Plugin corerun from marketplace corerun in corerun-ai/skills. Commands read from the repo's manifest — 1 month ago.

Skills (14)

• corerun-admin

  Administer a corerun organisation from the CLI -- people and their workspace roles, invitations, groups, identity providers (SSO), service accounts for CI, the licence, per-workspace quotas, shared git connections, usage and security settings. Use when asked to onboard or offboard someone, change wh

• corerun-agents

  Build, share and talk to agents on corerun — a coding-agent harness in a sandbox per person, with instructions, projects of files, connectors (MCP servers and Kubernetes clusters), custom MCP servers the platform runs (from API descriptions or an image), policies saying what it may do with them and

• corerun-datasets

  Find, import, upload and inspect datasets on corerun — from HuggingFace, Kaggle, a URL, or local files and folders, with schema, format, size, tags and licence. Use when a task needs training data, when asked what data is available, to put local data on the platform, or before wiring a dataset into

• corerun-devops

  Install, upgrade, back up, monitor and troubleshoot an on-premises corerun installation with Helm on Kubernetes -- the enterprise edition, its licence, first-run bootstrap, identity provider, object store, compute onboarding and secrets. Use when asked to deploy corerun on a customer's own cluster,

• corerun-endpoints

  Model endpoints on corerun — the stable address an application calls, its API keys, and publishing models behind it including ones corerun does not run. Use when asked for the URL to call a model, to rotate or find a key, to put a provider's model behind the platform's address, or when a client gets

• corerun-finetune

  Run LoRA, QLoRA and full fine-tuning jobs on corerun — size one against the compute before it runs, pick a base model and dataset, register the result in the model registry, track it to completion. Use when asked to fine-tune, adapt, or specialise a model on a dataset.

• corerun-genai

  Read and judge what an agent did on corerun — traces and their span trees, the conversations they group into, the judges that score them, review queues for human verdicts, issue detection over a sample of traces, and evaluation runs. Use when asked why an agent answered as it did, what it costs or h

• corerun-inference

  Deploy and manage model-serving endpoints on corerun — vLLM for language models, Triton for the rest, LoRA adapters, the publisher's recipe for each model. Use for any request to serve a model, get an OpenAI-compatible endpoint, change how one runs, move it behind another endpoint, or find out why a

• corerun-jobs

  Submit, monitor, and debug training jobs on corerun — run a folder of training code or a workspace code repository on a GPU profile, from flags or a corerun.yaml job file, list jobs, follow logs, check status, stop or delete them. Use for any request to train a model, run a training script, check wh

• corerun-models

  Work with the corerun model registry — publish catalogue models into it, list and create registered models, import from HuggingFace, push and pull versions, stage them through staging and production, and set aliases like champion. Use when asked to publish, register, import, promote, fetch or versio

• corerun-notebooks

  Start, find, share, move, stop and delete notebooks on corerun — interactive sessions on a cluster or bare-metal host, with GPUs, mounted datasets and environment variables — and read a stopped notebook's saved files. Use when asked for a notebook or a place to work interactively, for a notebook's U

• corerun-quota

  Check what the workspace has room for before allocating anything on corerun — GPUs, jobs, notebooks, inference servers, storage. Use before submitting a job, starting a fine-tune, or deploying an endpoint, and whenever a creation fails with a limit or quota error.

• corerun-tokens

  Issue, list and revoke corerun tokens for things that are not a person — an OpenTelemetry exporter sending traces, a CI job, a script — each limited to the scopes it needs. Use when something needs to call corerun without a person signed in, when setting up trace export, or when a token may have lea

• corerun-workspaces

  Create and delete corerun workspaces, choose which one the CLI acts in, and manage the organization's storage accounts and compute. Use when asked to set up a new workspace, switch between workspaces, add an S3 or ObjectIO account, or connect a cluster or bare-metal host.

Similar plugins