vastai-pack

Run a credential-safe disposable Vast.ai GPU CI job with a scoped key, immutable image, cost ceiling, terminal-state deadline, and guaranteed destruction. Use when GPU-only acceptance must run in CI. Trigger with: "run GPU tests in CI", "add Vast.ai to GitHub Actions", "make Vast.ai CI clean up".

Classify Vast.ai authentication, offer, instance-state, SSH, image, credit, and API failures before choosing a recovery action. Use when automation is stuck or a GPU workload cannot start. Trigger with: "diagnose Vast.ai", "why is my instance scheduling", "fix a Vast.ai CLI error".

Analyze and execute a checkpointed Vast.ai training job from offer policy through artifact recovery and destruction. Use when a repeatable single-job GPU run needs budget and interruption controls. Trigger with: "run training on Vast.ai", "checkpoint a Vast.ai job", "recover Vast.ai training artifacts".

Build and roll out a Vast.ai Serverless endpoint with measured autoscaling, a canary template, and a zero-downtime worker update. Use when production inference capacity or model bytes change. Trigger with: "deploy Vast.ai Serverless", "tune a Vast.ai worker group", "roll a model without downtime".

Reduce Vast.ai GPU, storage, and bandwidth spend without weakening workload requirements or leaving stopped resources billable. Use when selecting offers, setting spot policy, cleaning idle resources, or reconciling invoices. Trigger with: "optimize Vast.ai cost", "find Vast.ai cost leaks", "compare Vast.ai offers".

Manage datasets, checkpoints, models, and artifacts through Vast.ai instances, volumes, and cloud connections with integrity and teardown controls. Use when data must move to, from, or between Vast.ai resources. Trigger with: "copy data to Vast.ai", "protect Vast.ai checkpoints", "recover data before destroying an instance".

Collect a minimal, redacted Vast.ai diagnostic manifest that support can act on without exposing API keys, SSH material, workload secrets, or unnecessary customer data. Use when escalating a platform or workload fault. Trigger with: "build a Vast.ai debug bundle", "collect Vast.ai support evidence", "redact Vast.ai diagnostics".

Deploy an immutable service or batch worker to a Vast.ai instance with template, health, artifact, rollback, and cleanup controls. Use when shipping an instance-based workload rather than Serverless. Trigger with: "deploy to a Vast.ai instance", "release a Vast.ai template", "roll back a Vast.ai deployment".

Design and verify native Vast.ai Teams roles plus scoped API keys for least-privilege renter operations. Use when onboarding members, separating duties, or constraining automation. Trigger with: "configure Vast.ai RBAC", "create a Vast.ai team role", "scope a Vast.ai API key".

Rent, verify, use, and destroy a first Vast.ai GPU instance with bounded cost and explicit failure cleanup. Use when running a safe platform canary or onboarding exercise. Trigger with: "rent my first Vast.ai GPU", "run a Vast.ai canary", "test Vast.ai end to end".

Analyze, diagnose, and recover a Vast.ai renter workload from outbid, exited, offline, scheduling, low-credit, or data-risk incidents while controlling billing and destructive actions. Use when a Vast.ai workload or account enters an incident state. Trigger with: "Vast.ai incident", "recover an offline Vast.ai job", "stop emergency Vast.ai spend".

Install the supported Vast.ai CLI and establish a least-privilege authentication boundary without leaking or persisting the wrong key. Use when bootstrapping a workstation, CI runner, or automation account. Trigger with: "install Vast.ai", "configure VAST_API_KEY", "create a scoped Vast.ai key".

Build a fast local development loop that promotes a CPU-tested change to one disposable Vast.ai GPU canary and always tears it down. Use when iterating on images, startup scripts, or training code. Trigger with: "test locally then on Vast.ai", "build a GPU canary loop", "shorten Vast.ai iteration".

Migrate a GPU workload from another provider to Vast.ai through inventory, container and data parity, a checkpointed canary, measured comparison, and rollback. Use when planning or executing a provider cutover. Trigger with: "migrate Runpod to Vast.ai", "move GPU jobs to Vast.ai", "validate a Vast.ai migration".

Separate development, staging, and production Vast.ai identities, templates, labels, budgets, data, and evidence so one environment cannot mutate another. Use when establishing or auditing environment isolation. Trigger with: "set up Vast.ai environments", "separate Vast.ai prod and dev", "govern Vast.ai templates by environment".

Observe Vast.ai renter instances and Serverless resources using actionable state, logs, queue, utilization, balance, and cost signals. Use when building dashboards, alerts, or release telemetry. Trigger with: "monitor Vast.ai", "alert on Vast.ai instances", "observe Vast.ai Serverless".

Tune Vast.ai offer selection, startup, data feeding, GPU memory, and scaling from measured useful-work throughput. Use when jobs are slow, GPUs are underutilized, or faster offers cost more. Trigger with: "tune Vast.ai performance", "improve GPU utilization", "compare Vast.ai throughput".

Analyze readiness and issue a production go/no-go decision for a Vast.ai renter workload using retained evidence for cost, capacity, security, recovery, observability, and teardown. Use when a release needs production approval. Trigger with: "approve Vast.ai production", "run the Vast.ai launch checklist", "is this Vast.ai workload production ready".

Diagnose and design bounded Vast.ai CLI and REST traffic around per-endpoint, per-identity rate limits and the absence of Retry-After headers. Use when implementing polling, batch automation, or 429 recovery. Trigger with: "handle Vast.ai 429", "reduce Vast.ai polling", "set Vast.ai retry policy".

Design a governed Vast.ai GPU control plane that separates planning, paid mutation, execution, recovery, evidence, and teardown. Use when reviewing a production architecture spanning instances or Serverless. Trigger with: "design a Vast.ai architecture", "govern GPU workload lifecycles", "review a Vast.ai platform".

Choose the correct Vast.ai Python client and implement typed, bounded, ownership-safe GPU lifecycles. Use when integrating the high-level SDK, SyncClient, AsyncClient, or Serverless client. Trigger with: "use the Vast.ai SDK", "wrap a Vast.ai instance lifecycle", "choose SyncClient or AsyncClient".

Harden Vast.ai renter access across scoped API keys, SSH, images, hosts, data, and teardown. Use when reviewing security before placing workloads on marketplace hardware. Trigger with: "secure Vast.ai", "review Vast.ai credentials", "protect data on a rented GPU".

Upgrade or roll back the Vast.ai CLI, Python SDK import surface, and workload template without speculative API-version changes. Use when changing a client pin, SDK import, or production template. Trigger with: "upgrade Vast.ai CLI", "migrate from vastai_sdk", "roll a Vast.ai template update".

Analyze, verify, and operate Vast.ai notification webhooks with exact-body HMAC, replay protection, deduplication, durable enqueue, and failure-aware acknowledgment. Use when notification events must drive automation. Trigger with: "receive Vast.ai webhooks", "verify X-Vast-Signature-256", "handle duplicate Vast.ai events".