feat(asr): support additional Bailian speech models

[LauraGPT]

User description

Closes #850

Summary

• route Bailian model families to classic realtime, Qwen realtime, synchronous multimodal, or asynchronous transcription clients
• support Fun-ASR, Fun-ASR-MTL, Paraformer file transcription, Qwen3-ASR Flash/filetrans, and 8 kHz Fun-ASR realtime models
• align desktop and mobile routing, provider validation, model hints, localized UI, and documentation
• implement temporary OSS upload, async task polling, model-specific response parsing, cancellation-aware QA flows, and dynamic long-recording timeouts
• harden credential and anonymous requests against redirects, untrusted transfer URLs, SSRF, and stalled requests

Verification

• cargo check --tests
• cargo test --lib: 773 passed (serial full-suite run)
• npm run build: 2445 modules transformed
• npm test: 26 frontend tests passed
• git diff --check

Notes

DashScope temporary OSS upload is documented as a development and low-concurrency path. The added documentation calls out that limitation for users.

---

PR Type

Enhancement, Tests

---

Description

• Extend Bailian ASR model routing to all supported families

• Implement async transcription protocol with OSS upload and polling

• Add 8kHz PCM downsample and chunk size for 8k realtime models

• Align desktop and mobile routing, model hints, and validation

---

Diagram Walkthrough

flowchart LR
  model["Model Name"] --> classify{"Model Prefix"}
  classify -->|"fun-asr-realtime* / paraformer-realtime* / sensevoice-realtime* / fun-asr-flash-8k-realtime*"| ws["Classic WebSocket (bailian.rs)"]
  classify -->|"qwen3-asr-flash-realtime*"| qwen["Qwen Realtime WebSocket (qwen_realtime.rs)"]
  classify -->|"fun-asr-flash* / qwen3-asr-flash"| sync["Synchronous Multimodal (dashscope_multimodal.rs)"]
  classify -->|"fun-asr / fun-asr-mtl* / paraformer* / qwen3-asr-flash-filetrans*"| async["Async Transcription (dashscope_multimodal.rs)"]

Loading

File Walkthrough

	Relevant files
Enhancement	6 files bailian.rs Add 8kHz downsample and chunk size                                              +131/-7  dashscope_multimodal.rs Implement async transcription protocol                                      +677/-11 providers.rs Extend model list and validation                                                  +93/-36  coordinator.rs Extend model routing and endpoint protocols                            +51/-30  dictation.rs Use dynamic timeout for multimodal ASR                                      +1/-1      ProvidersSection.tsx Classify model protocol for UI hints                                          +14/-6
Bug fix	1 files qa_session.rs Add cancel support for DashScope multimodal                            +34/-21
Documentation	6 files en.ts Update model hint labels for sync/async                                    +2/-1      ja.ts Update model hint labels for sync/async                                    +2/-1      ko.ts Update model hint labels for sync/async                                    +2/-1      zh-CN.ts Update model hint labels for sync/async                                    +2/-1      zh-TW.ts Update model hint labels for sync/async                                    +2/-1      bailian-asr-models.md Document Bailian ASR model support                                              +32/-0

---

[github-actions]

PR Reviewer Guide 🔍

(Review updated until commit a58630b)

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶 850 - Partially compliant Compliant requirements: 用户手填 fun-asr、qwen3-asr-flash、qwen3-asr-flash-filetrans 等官方模型时，不再被 resolve_effective_asr_provider 直接拒绝 每种新协议有对应的 Rust 客户端 + coordinator 分支 + 凭据校验 设置页模型提示能区分：实时 WebSocket / 同步录音文件 / 异步文件转写 现有已支持模型（fun-asr-realtime、qwen3-asr-flash-realtime*、fun-asr-flash-2026-06-15）行为不回归 补充单元测试覆盖模型名 → 协议路由表 Non-compliant requirements: "" Requires further human verification: ""

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ No major issues detected

[LauraGPT]

Follow-up on the PR-Agent 8 kHz tail note: the decimator intentionally keeps source indices 0, 2, 4, ... . Therefore, when a stream ends after three complete 16 kHz samples, the buffered third sample is the next kept 8 kHz sample rather than half of an output pair.

Signed commit a58630b extracts that tail flush into a named helper and adds a regression proving [1, 2, 3] -> [1, 3] at end of stream. Fresh verification passed cargo check --tests and all 773 Rust tests serially. (Two unrelated localhost connection-pool tests raced in one parallel run; both passed in isolation, and the full serial run is 773/773.)

[github-actions]

Persistent review updated to latest commit a58630b