chore(deps): bump @qvac/sdk from 0.13.5 to 0.18.2

[dependabot]

Bumps @qvac/sdk from 0.13.5 to 0.18.2.

Release notes

Sourced from @​qvac/sdk's releases.

QVAC SDK v0.18.2

📦 NPM: https://www.npmjs.com/package/@​qvac/sdk/v/0.18.2

QVAC SDK 0.18.2 bumps @qvac/diffusion-cpp to ^0.18.0.

QVAC SDK v0.18.1

📦 NPM: https://www.npmjs.com/package/@​qvac/sdk/v/0.18.1

QVAC SDK 0.18.1 adds human-readable descriptions on every llamacpp modelConfig field and exports those schemas from @qvac/sdk/schemas. CLI, typed-client, and Python generators can now surface what each completion and embedding option means. The CosyVoice3 companion-set cache key also changes, so the first load after upgrade uses a new companion cache folder. Load APIs are otherwise unchanged.

Table of Contents

• New APIs
  • Config schemas at @qvac/sdk/schemas
• Bug Fixes
  • CosyVoice3 companion cache folder

New APIs

Config schemas at @qvac/sdk/schemas

@qvac/sdk/schemas exports llamacppCompletionConfigSchema, llamacppEmbeddingConfigSchema, and modelSourceSchema. Field .describe() text comes from the addon README / index.d.ts (or an existing SDK description) and is written into contract/schema.json. The same descriptions land on the generated Python pydantic fields.

import { llamacppCompletionConfigSchema } from '@qvac/sdk/schemas'
llamacppCompletionConfigSchema.shape.ctx_size.description
// "Context window size in tokens; 0 uses the model's trained context length. Default 1024."

The internal schema identifiers are unchanged.

Bug Fixes

CosyVoice3 companion cache folder

CosyVoice3 companion files still download with the LLM, as in 0.18.0. The companion-set cache key for TTS_COSYVOICE3_LLM_COSYVOICE_Q8_0 changed, so the first load after upgrading fills a new cache folder. Later loads reuse that folder. Speech APIs and pace / instruct rules are unchanged.

QVAC SDK v0.18.0

📦 NPM: https://www.npmjs.com/package/@​qvac/sdk/v/0.18.0

QVAC SDK 0.18.0 adds VisionPsy Nano multimodal constants, Audio8 and CosyVoice3 TTS engines, Indic Conformer transcription, and ACE-Step cover generation from a source track. One loaded LLM can now serve several completions at once, sharded GGUFs load directly from disk, and loadModel can fall back to a backup source when the origin fails. The dynamic toolsMode config is removed, and CosyVoice3 pace is restricted to slow | moderate | fast.

Table of Contents

• Breaking Changes
  • Dynamic toolsMode Removed
  • CosyVoice3 Pace Values
• New APIs

... (truncated)

Changelog

Sourced from @​qvac/sdk's changelog.

[0.18.2]

📦 NPM: https://www.npmjs.com/package/@​qvac/sdk/v/0.18.2

QVAC SDK 0.18.2 bumps @qvac/diffusion-cpp to ^0.18.0.

[0.18.1]

📦 NPM: https://www.npmjs.com/package/@​qvac/sdk/v/0.18.1

QVAC SDK 0.18.1 adds human-readable descriptions on every llamacpp modelConfig field and exports those schemas from @qvac/sdk/schemas. CLI, typed-client, and Python generators can now surface what each completion and embedding option means. The CosyVoice3 companion-set cache key also changes, so the first load after upgrade uses a new companion cache folder. Load APIs are otherwise unchanged.

New APIs

Config schemas at @qvac/sdk/schemas

@qvac/sdk/schemas exports llamacppCompletionConfigSchema, llamacppEmbeddingConfigSchema, and modelSourceSchema. Field .describe() text comes from the addon README / index.d.ts (or an existing SDK description) and is written into contract/schema.json. The same descriptions land on the generated Python pydantic fields.

import { llamacppCompletionConfigSchema } from '@qvac/sdk/schemas'
llamacppCompletionConfigSchema.shape.ctx_size.description
// "Context window size in tokens; 0 uses the model's trained context length. Default 1024."

The internal schema identifiers are unchanged.

Bug Fixes

CosyVoice3 companion cache folder

CosyVoice3 companion files still download with the LLM, as in 0.18.0. The companion-set cache key for TTS_COSYVOICE3_LLM_COSYVOICE_Q8_0 changed, so the first load after upgrading fills a new cache folder. Later loads reuse that folder. Speech APIs and pace / instruct rules are unchanged.

[0.18.0]

📦 NPM: https://www.npmjs.com/package/@​qvac/sdk/v/0.18.0

QVAC SDK 0.18.0 adds VisionPsy Nano multimodal constants, Audio8 and CosyVoice3 TTS engines, Indic Conformer transcription, and ACE-Step cover generation from a source track. One loaded LLM can now serve several completions at once, sharded GGUFs load directly from disk, and loadModel can fall back to a backup source when the origin fails. The dynamic toolsMode config is removed, and CosyVoice3 pace is restricted to slow | moderate | fast.

Breaking Changes

Dynamic toolsMode Removed

TOOLS_MODE, ToolsMode, and the toolsMode load-config field are gone. Tools are always prepended after the system message (the previous static default). Passing toolsMode to loadModel now fails validation instead of being ignored. Drop the key from existing configs.

Before:

import { loadModel, TOOLS_MODE } from '@qvac/sdk'
</tr></table>

... (truncated)

Commits

• 2ce4ed8 QVAC-24089 test: skip redundant CosyVoice3 e2e tests on mobile (#4030)
• 277836d fix[notask]: align SDK @​qvac/inference range to the 0.18 line (#4021)
• 1d79313 chore[bc]: remove langdetect-text-cld2 from the monorepo (#4011)
• 4994117 QVAC-21751 chore[bc]: use @​qvac/inference for SDK (#3595)
• 2b049b8 QVAC-23810 doc: cover multi-job continuous batching and per-request cancel (#...
• ae97a0e QVAC-23933 chore[skiplog]: backmerge release-sdk-0.18.1 — version bump, chang...
• 86c3f0d QVAC-23581 feat[api]: add LTX IC-LoRA reference conditioning to video generat...
• cf5fd12 doc[notask]: restore SDK TSDoc completeness gate (#3510)
• 2a2ffe9 QVAC-22963 chore[skiplog]: backmerge release-sdk-0.18.0 — version bump, chang...
• c9247f7 QVAC-23933 feat[api]: describe llamacpp modelConfig fields and export config ...
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)