[Bugfix] Restore MiniCPM duplex Seed-TTS startup

[NolenLiang]

Purpose

Fixes #8138. Since #7974, native duplex Seed-TTS can remain listening without starting its first response. Mark nonempty initial text as pending input during session initialization; existing generation logic consumes it while preserving forced listening and post-response silence.

Keep #7974's discarded-lookahead guard. Source inspection and synthetic-logits replay indicate that earlier discarded sampling could clear current_turn_ended, masking the missing seed initialization. No historical GPU token/state trace was captured.

Test Plan

vLLM Version: 0.30.0.

vLLM-Omni Commit: 6fb2b36 before/after the patch; b308e19 + patch; pristine controls d486bf0 and 7a8d956.

One GB200; openbmb/MiniCPM-o-4_5 at 503e754, zhaochenyang20/seed-tts-eval at 8f5e1aa. Retain the original deployment, four English targets, two warmups, assertions and timeouts; model/dataset paths point to local copies. CI uses H100. Benchmark command from CI, followed by focused checks:

export BENCHMARK_DIR=tests/dfx/perf/results
pytest -s -v tests/dfx/perf/scripts/run_benchmark.py \
  --test-config-file tests/dfx/perf/tests/test_minicpmo_4_5_duplex_seed_tts.json \
  -m "H100 and B200 and cards_1"
pytest -q tests/worker/test_native_duplex_hooks.py
pytest -s -v tests/e2e/online_serving/test_minicpmo_4_5_duplex.py \
  -k "seeded or sequential_sessions_are_independent" \
  -m "H100 and cards_1" --run-level advanced_model

Test Result

• Original benchmark: d486bf0 passes; its child 7a8d956 and issue revision 6fb2b36 fail in both warmups and the measured request. 6fb2b36 + patch completes 1/1 request and four audio targets.
• Eight new regression cases: baseline 3 fail, 5 pass; patched hooks test file 91 pass, including on b308e19. Coverage includes empty/missing text, forced listening and post-response silence. Changed-file pre-commit checks pass with the repository-pinned tools.
• Existing seeded E2E: 5 pass, 1 fail. The remaining test_duplex_seeded_text_to_text_needs_no_reference_voice case was also run in isolation:

Source	Isolated no-reference text result
d486bf0	Pass
7a8d956	Fail: 12 listen events, no response.done
b308e19 + patch	Same failure

The controlled comparisons use the same GPU, runtime, model and inputs; comparability checks pass. This confirms the text-only regression at #7974, but the patch leaves it unresolved. Its remaining internal cause is not isolated. The demonstrated repair is Seed-TTS startup, not the entire Duplex Test step or #8123.

Nightly 16046 already passed Seed-TTS on vLLM 0.30.0 before #7974; merge build 16048 used the same version and failed all five seeded cases. This boundary is reproduced locally above.

AI assistance: Codex assisted with testing and drafting; Claude performed an independent review.

[vllm-omni-review-bot]

This PR appears to belong to: docs/design/module/model_integration.md, docs/design/module/ar_runtime.md.

Module owners: @tzhouam @fake0fan @Gaohan123

Routing: @tzhouam via module of the changed files, CODEOWNERS; @fake0fan via module of the changed files; @Gaohan123 via module of the changed files

@NolenLiang, please review your own changes and leave a short self-review comment describing what you checked. PRs without author self-review may not be assigned a reviewer.

Please take a look when you have a chance. If you would like an automated review, mention @vllm-omni-review-bot in a comment.

[vllm-omni-review-bot]

Omni ReviewBot triage note

Automated triage of commit c0cc95db19a0 produced:

• Priority: high. Prompt maintainer attention is suggested.

These are automated triage suggestions only — the final decision belongs to the maintainers.

[NolenLiang]

Self review: Passed multiple repeated experiments, minimal fix is done.

[Gaohan123]

LGTM. Thanks