feat: Python version constraints, GpuType migration, idle_timeout fixes

[deanq]

Summary

• Document Python version constraints for deployed workers (GPU pinned to 3.12, CPU supports 3.10-3.12)
• Migrate GpuGroup to explicit GpuType values in example worker files (e.g. GpuGroup.ADA_24 -> GpuType.NVIDIA_GEFORCE_RTX_4090)
• Fix idle_timeout values and units (parameter is seconds, not minutes)
• Correct GpuType enum names in docs to match actual SDK values

Changes

1. Python version documentation

• Added "Python version in deployed workers" section to README
• Set requires-python to >=3.10,<3.13 and runpod-flash to >=1.7.0

2. GpuGroup -> GpuType migration in worker files

• Replaced GpuGroup.ADA_24 with GpuType.NVIDIA_GEFORCE_RTX_4090 in hello_world example
• Note: GpuGroup.ANY and GpuGroup are intentionally kept in docs for troubleshooting and flexible GPU selection contexts (they remain valid SDK APIs)

3. GpuType naming corrections in docs

• Fixed GpuType.NVIDIA_A100_80GB -> GpuType.NVIDIA_A100_80GB_PCIe (correct enum name)
• Fixed GpuType.NVIDIA_H100_80GB -> GpuType.NVIDIA_H100_80GB_HBM3 (correct enum name)
• Changed "GPU unavailable" quick fix to recommend GpuGroup.ANY or GpuType.ANY

4. idle_timeout and dependency fixes

• Corrected idle_timeout units in README tables from minutes to seconds
• Fixed config dict idleTimeout values in autoscaling workers to match decorator values
• Changed README example idle_timeout=1 to idle_timeout=300 for consistency
• Removed unnecessary dependency pins, added missing accelerate dependency

Test plan

• flash run starts successfully with updated examples
• All worker files import without errors
• Verify idle_timeout config dict values match decorator values
• Confirm all GpuType references use valid enum names from SDK

[Copilot]

Pull request overview

Updates the Flash examples/docs to reflect current runtime constraints (Python versions), standardize GPU selection guidance (favoring explicit GpuType values), and correct autoscaling idle_timeout configuration to use seconds.

Changes:

• Document deployed-worker Python version constraints and tighten requires-python while bumping runpod-flash minimum version.
• Replace many GpuGroup.ANY usages with explicit GpuType values in docs and examples.
• Convert idle_timeout examples/docs to seconds, remove unnecessary defaults in simple examples, and relax dependency pins for broader Python compatibility.

Reviewed changes

Copilot reviewed 22 out of 22 changed files in this pull request and generated 13 comments.

Show a summary per file

File	Description
pyproject.toml	Constrains supported Python versions and bumps runpod-flash minimum version.
docs/cli/workflows.md	Updates workflow snippets for GpuType usage and corrects idle_timeout examples.
docs/cli/troubleshooting.md	Updates troubleshooting guidance to reference GpuType/GpuGroup and revised GPU availability guidance.
README.md	Adds deployed-worker Python version notes and updates idle_timeout unit wording.
CONTRIBUTING.md	Migrates contributing examples from GpuGroup.ANY to explicit GpuType usage.
CLAUDE.md	Updates internal guide examples and symbol table to reflect GpuType usage.
05_data_workflows/01_network_volumes/gpu_worker.py	Switches to GpuType, fixes idle_timeout units, and adds missing deps for the SD example.
05_data_workflows/01_network_volumes/cpu_worker.py	Adds explicit idle_timeout for CPU worker.
04_scaling_performance/01_autoscaling/gpu_worker.py	Migrates to GpuType and converts scale-to-zero timeout to seconds.
04_scaling_performance/01_autoscaling/cpu_worker.py	Converts CPU scale-to-zero timeout to seconds.
04_scaling_performance/01_autoscaling/README.md	Updates autoscaling reference/examples to match GpuType and new timeout semantics.
03_advanced_workers/05_load_balancer/gpu_lb.py	Migrates GPU LB example to explicit GpuType and expands config for readability.
03_advanced_workers/05_load_balancer/cpu_lb.py	Adds explicit workers to CPU LB endpoint configuration.
03_advanced_workers/05_load_balancer/README.md	Updates load balancer docs to use GpuType in examples.
02_ml_inference/01_text_to_speech/gpu_worker.py	Converts idle_timeout to seconds for GPU worker example.
01_getting_started/04_dependencies/cpu_worker.py	Unpins pandas/numpy to improve cross-Python compatibility.
01_getting_started/04_dependencies/README.md	Updates dependency guidance from strict pinning to constraints-based recommendations.
01_getting_started/03_mixed_workers/gpu_worker.py	Removes unnecessary idle_timeout default from a simple GPU endpoint.
01_getting_started/03_mixed_workers/README.md	Converts idle_timeout examples from minutes to seconds.
01_getting_started/02_cpu_worker/cpu_worker.py	Removes unnecessary workers/idle_timeout defaults from the simplest CPU example.
01_getting_started/01_hello_world/gpu_worker.py	Changes default GPU selection away from GpuGroup.ANY.
01_getting_started/01_hello_world/README.md	Updates examples to use GpuType and refreshes guidance for choosing GPU types.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

[Copilot]

Pull request overview

Copilot reviewed 22 out of 22 changed files in this pull request and generated 8 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.