build: remove aiconfigurator-core compatibility wheel

[jasonqinzhou]

Summary

• remove the payload-free aiconfigurator-core compatibility project
• stop building and installing a second Python wheel in the release Dockerfile
• keep the self-contained aiconfigurator wheel as the only Python distribution
• preserve the rust/aiconfigurator-core crate and the bundled aiconfigurator_core PyO3 extension
• update wheel verification and architecture/install documentation for the one-wheel layout

Why

Current main already puts the CLI, SDK, data, Spica, and native extension in the aiconfigurator wheel. The separate aiconfigurator-core wheel contains no payload; it only installs the exact matching aiconfigurator version, so it adds a release surface without providing an independently useful package.

This intentionally keeps the current single-wheel ownership model instead of advancing the standalone-wheel experiments in #1322 and #1327.

Impact

• pip install aiconfigurator and import aiconfigurator_core continue to work.
• Rust consumers keep the aiconfigurator-core crate unchanged.
• Release builds now produce only aiconfigurator-*.whl.
• Accuracy regression tests retain their conditional install path for historical split-era baselines.

Validation

• built the Maturin wheel locally; output contained exactly one aiconfigurator-0.10.0-*.whl
• python3 tools/verify_release_wheels.py <dist>
• isolated no-dependency wheel install followed by import aiconfigurator and import aiconfigurator_core
• cargo check -p aiconfigurator-core
• Ruff checks for the modified Python files
• 16 passed in tests/unit/sdk/test_package_layering.py
• Developer Universe JavaScript syntax check

Docker is not installed in the local environment; the repository's Build/Test CI exercises the updated Dockerfile.

Summary by CodeRabbit

• New Features

  • Release packaging now produces a single self-contained AIC wheel and no longer includes the legacy compatibility package.
  • Build verification now checks for the self-contained release artifact only.

• Bug Fixes

  • Prevented packaging from accidentally including the removed compatibility wheel.
  • Updated release checks to fail if the legacy core wheel is present.

• Documentation

  • Refreshed install, build, and architecture notes to match the new wheel format.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Walkthrough

This PR removes the aiconfigurator-core compatibility metapackage, deletes its pyproject.toml, updates release wheel verification to reject core wheel dependency, narrows Docker runtime packaging to a single self-contained wheel, and updates README, workflow comments, and the universe map docs to match.

Changes

aiconfigurator-core removal and wheel verification

Layer / File(s)	Summary
Remove package config src/aiconfigurator-core/pyproject.toml	Deletes the entire aiconfigurator-core package configuration.
Verification script rejects core wheel tools/verify_release_wheels.py	_verify_main_wheel now returns None, rejects wheels depending on aiconfigurator-core, requires Version metadata; main() rejects any aiconfigurator_core-*.whl and verifies only the main wheel.
Docker build/runtime packaging docker/Dockerfile	Removes the separate core wheel build step and narrows runtime install to aiconfigurator-*.whl only; updates build comment.
README and workflow docs README.md, .github/workflows/accuracy-regression-testing.yml	Updates install/build instructions to state no separate core distribution is needed and clarifies old-revision dependency install comment.

Universe map wheel node rename

Layer / File(s)	Summary
Rename core_wheel to aic_wheel docs/universe/index.html	Renames the wheel node/layer label, edges, group edges, note map, TODOS key, source repos, and overview text from core_wheel/aiconfigurator-core to aic_wheel/aiconfigurator-main.

Estimated code review effort: 2 (Simple) | ~12 minutes

Poem

A core wheel packed up and rolled away,
One self-contained wheel now saves the day. 🛞
Docs, docker, and maps all realigned,
No stray metapackage left behind.
Hop, hop — ship it clean today! 🐇

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title is concise and accurately summarizes the main change: removing the aiconfigurator-core compatibility wheel.
Description check	✅ Passed	The description is detailed and covers overview, rationale, impact, and validation, but it does not follow the template sections or include reviewer start and related issue info.

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

AIC Accuracy Regression Testing

Workflow result: success

Old revision: 0828d6b7e4a7880079443b1c6f9c148d85bdbf54

Summary

• # samples added: 0
• ttft_mape_improvement_%: 0.000000
• tpot_mape_improvement_%: 0.000000

Comparison summary table

partition	num_samples_added	ttft_mape_improvement_%	tpot_mape_improvement_%
all	0	0.000000	0.000000
agg	0	0.000000	0.000000
disagg	0	0.000000	0.000000
MiniMax-M2.5	0	0.000000	0.000000
Qwen3.5-397B-A17B	0	0.000000	0.000000
Kimi-K2.5	0	0.000000	0.000000
DeepSeek-V3	0	0.000000	0.000000
DeepSeek-R1-NVFP4	0	0.000000	0.000000
DeepSeek-V3.1-NVFP4	0	0.000000	0.000000
Qwen3-32B-NVFP4	0	0.000000	0.000000
DeepSeek-R1	0	0.000000	0.000000
Qwen3-32B	0	0.000000	0.000000
Qwen3-235B-A22B-FP8	0	0.000000	0.000000
Llama-3.1-8B-Instruct-FP8	0	0.000000	0.000000
b300_sxm	0	0.000000	0.000000
gb200	0	0.000000	0.000000
b200_sxm	0	0.000000	0.000000
gb300	0	0.000000	0.000000
h200_sxm	0	0.000000	0.000000
h100_sxm	0	0.000000	0.000000
vllm-agg	0	0.000000	0.000000
sglang-agg	0	0.000000	0.000000
trtllm-disagg	0	0.000000	0.000000
sglang-disagg	0	0.000000	0.000000
trtllm-agg	0	0.000000	0.000000
vllm-disagg	0	0.000000	0.000000
MiniMax-M2.5|b300_sxm|vllm|agg	0	0.000000	0.000000
Qwen3.5-397B-A17B|b300_sxm|sglang|agg	0	0.000000	0.000000
Kimi-K2.5|gb200|trtllm|disagg	0	0.000000	0.000000
Kimi-K2.5|b200_sxm|vllm|agg	0	0.000000	0.000000
Kimi-K2.5|b300_sxm|vllm|agg	0	0.000000	0.000000
MiniMax-M2.5|b200_sxm|vllm|agg	0	0.000000	0.000000
DeepSeek-V3|gb200|trtllm|disagg	0	0.000000	0.000000
DeepSeek-V3|gb300|trtllm|disagg	0	0.000000	0.000000
DeepSeek-V3|gb200|sglang|disagg	0	0.000000	0.000000
DeepSeek-V3|gb300|sglang|disagg	0	0.000000	0.000000
DeepSeek-V3|h200_sxm|sglang|agg	0	0.000000	0.000000
DeepSeek-V3|b200_sxm|trtllm|disagg	0	0.000000	0.000000
DeepSeek-V3|h200_sxm|sglang|disagg	0	0.000000	0.000000
MiniMax-M2.5|h200_sxm|vllm|agg	0	0.000000	0.000000
Qwen3.5-397B-A17B|b200_sxm|sglang|agg	0	0.000000	0.000000
DeepSeek-V3|b200_sxm|sglang|disagg	0	0.000000	0.000000
DeepSeek-V3|b200_sxm|sglang|agg	0	0.000000	0.000000
DeepSeek-V3|b200_sxm|trtllm|agg	0	0.000000	0.000000
DeepSeek-V3|h200_sxm|trtllm|disagg	0	0.000000	0.000000
DeepSeek-V3|h100_sxm|trtllm|disagg	0	0.000000	0.000000
DeepSeek-V3|h200_sxm|trtllm|agg	0	0.000000	0.000000
Kimi-K2.5|gb200|vllm|disagg	0	0.000000	0.000000
DeepSeek-V3|b300_sxm|sglang|agg	0	0.000000	0.000000
DeepSeek-R1-NVFP4|b200_sxm|trtllm|agg	0	0.000000	0.000000
DeepSeek-V3.1-NVFP4|b200_sxm|trtllm|agg	0	0.000000	0.000000
Qwen3-32B-NVFP4|b200_sxm|trtllm|disagg	0	0.000000	0.000000
Qwen3-32B-NVFP4|b200_sxm|trtllm|agg	0	0.000000	0.000000
DeepSeek-R1|gb200|trtllm|agg	0	0.000000	0.000000
Qwen3-32B|gb200|trtllm|agg	0	0.000000	0.000000
DeepSeek-V3.1-NVFP4|gb200|trtllm|agg	0	0.000000	0.000000
Qwen3-32B-NVFP4|gb200|trtllm|disagg	0	0.000000	0.000000
Qwen3-32B-NVFP4|gb200|trtllm|agg	0	0.000000	0.000000
Qwen3-32B|h100_sxm|sglang|agg	0	0.000000	0.000000
Qwen3-32B|h100_sxm|sglang|disagg	0	0.000000	0.000000
Qwen3-235B-A22B-FP8|h100_sxm|trtllm|agg	0	0.000000	0.000000
Qwen3-32B|h100_sxm|trtllm|agg	0	0.000000	0.000000
Qwen3-32B|h100_sxm|vllm|agg	0	0.000000	0.000000
Qwen3-235B-A22B-FP8|h200_sxm|sglang|agg	0	0.000000	0.000000
Qwen3-235B-A22B-FP8|h200_sxm|vllm|agg	0	0.000000	0.000000
Llama-3.1-8B-Instruct-FP8|h200_sxm|vllm|agg	0	0.000000	0.000000
MiniMax-M2.5|h100_sxm|vllm|agg	0	0.000000	0.000000
DeepSeek-R1|b200_sxm|sglang|agg	0	0.000000	0.000000
DeepSeek-R1|h200_sxm|sglang|agg	0	0.000000	0.000000
DeepSeek-R1|b200_sxm|trtllm|agg	0	0.000000	0.000000
DeepSeek-R1|h200_sxm|trtllm|agg	0	0.000000	0.000000

Comparison plot

Artifact bundle: accuracy-regression-testing-results