Proposal: Jungle Grid bundle for agent-driven AI workload execution

[dejaguarkyng]

Hi Langflow maintainers,

I am preparing a Jungle Grid provider bundle for Langflow.

Jungle Grid is an execution layer for AI workloads and agents. A workflow submits workload intent — such as inference, training, fine-tuning, or batch execution — and Jungle Grid handles routing across compatible GPU capacity, managed execution, lifecycle monitoring, logs, cancellation, and output artifacts.

The proposed Langflow bundle exposes independent, composable components:

• Estimate Job — preview workload classification, expected capacity/routing, queue/start estimates, and cost estimates before execution.
• Submit Job — queue a real Jungle Grid workload.
• Get Job Status — retrieve job lifecycle status and details.
• Get Job Runtime — retrieve runtime information where the execution path exposes it.
• Get Job Logs — retrieve bounded stored logs for monitoring and debugging.
• Cancel Job — explicitly stop non-terminal work.
• List Job Artifacts — retrieve output artifacts from completed jobs.
• Create Artifact Download URL — mint a temporary artifact download URL.

Why this fits Langflow:

• Langflow agent workflows can trigger real asynchronous AI execution outside normal in-flow request limits.
• Agents can estimate before committing spend.
• The flow remains visually composable: estimate → submit → monitor → retrieve outputs.
• The implementation uses Jungle Grid’s documented public REST API with scoped bearer API keys.
• Side-effecting actions are separated and clearly labeled.
• The contribution will include tests, frontend bundle registration/icon wiring, documentation, and live validation using a disposable scoped Jungle Grid API key.

Jungle Grid documentation:

• https://junglegrid.dev/docs
• https://junglegrid.dev/docs/api
• https://junglegrid.dev/docs/mcp

I plan to open a PR against the active Langflow release candidate branch after completing the implementation and tests. Feedback on scope or component design is welcome.