Skip to content

feat(agentkit): add Engine v2.14 support to the Python SDK - #106

Open
seymourtang wants to merge 29 commits into
mainfrom
engine-v2.14
Open

seymourtang wants to merge 29 commits into
mainfrom
engine-v2.14

Conversation

@seymourtang

@seymourtang seymourtang commented Oct 9, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

Update the generated Python SDK and AgentKit layer to match the Engine v2.14 API, including RTZR speech recognition, the latest Sarvam TTS options, speak sentence-segmentation settings, and production Gemini TTS. Existing Gemini TTS preview integrations remain source-compatible while moving to the regional production gateway.

What Changed

  • Updated generated RTZR ASR, speak, Deepgram, Sarvam, and Gemini TTS models for the Engine v2.14 contract.

  • Added global-only RtzrSTT, generated aliases, and SpeakConfig/ParametersSpeak support for parameters.speak.batch.

  • Updated SarvamTTSOptions with speech sample rate, preprocessing, model, flattened additional_params, and compatibility for the deprecated sample-rate alias.

  • Aligned generated Deepgram credentials with asr.params.api_key and updated preset inference, legacy credential compatibility, vendor routing, and avatar sample-rate handling.

  • Promoted Gemini TTS from preview to production routing. Synchronous and asynchronous GeminiTTS sessions and raw tts.vendor = "gemini" configurations use the configured regional endpoint for the complete lifecycle without agora-feature: gemini-live.

  • Registered Gemini TTS in the production global vendor catalog while retaining package-root, agentkit.preview, and agentkit.preview.gemini_tts imports as aliases. The v2.11.0 model, defaults, validation, additional_params, and skip_patterns remain supported; named fields override passthrough values.

  • Kept generated style optional. Historical preview calls that omit it continue to serialize without a style field.

  • Updated AgentKit documentation, regional routing guidance, changelog entries, and regression coverage.

Testing

  • Focused Gemini TTS, preview compatibility, and generated schema tests: 61 passed
  • Full custom test suite: 456 passed, 1 skipped
  • Coverage includes synchronous and asynchronous US/EU/AP production routing, raw configurations, complete lifecycle requests, catalog registration, preview import compatibility, additional-parameter precedence, and skip patterns.

Anam Avatar portrait options

  • Promoted Engine 2.11+ Anam avatar_model, video_width, and video_height parameters to first-class AgentKit options.
  • Supports Cara 4 portrait mode with cara_mk4; width and height must be supplied together (for example, 720x1280), or both may be omitted to keep the model default.
  • Explicit options override matching additional_params values.

Additional validation

  • Focused Anam, vendor, and session tests: 52 passed
  • Ruff checks passed for changed implementation and test logic (excluding pre-existing import-order warnings)

fern-api Bot and others added 29 commits September 22, 2026 07:54
Generated by Fern
CLI Version: unknown
Generators:
  - fernapi/fern-python-sdk: 4.37.0
Patches applied (3):
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings
Generated by Fern
CLI Version: unknown
Generators:
  - fernapi/fern-python-sdk: 4.37.0
Patches applied (3):
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings

Patches with unresolved conflicts (1):
  - patch-ec343a2b: feat(agentkit): add Gemini TTS preview for v2.11.0
    Run `fern-replay resolve` to apply these customizations.
Generated by Fern
CLI Version: unknown
Generators:
  - fernapi/fern-python-sdk: 4.37.0
Patches applied (3):
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings
Patches replayed:
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings
  - patch-ec343a2b: feat(agentkit): add Gemini TTS preview for v2.11.0
Generated by Fern
CLI Version: unknown
Generators:
  - fernapi/fern-python-sdk: 4.37.0
Patches applied (3):
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings

Patches with unresolved conflicts (2):
  - patch-ec343a2b: feat(agentkit): add Gemini TTS preview for v2.11.0
  - patch-3768b9bf: feat(agentkit): add Tavus and Protoface aliases for v2.12.0
    Run `fern-replay resolve` to apply these customizations.
# Conflicts:
#	.fern/metadata.json
#	.fern/replay.lock
Generated by Fern
CLI Version: unknown
Generators:
  - fernapi/fern-python-sdk: 4.37.0
Patches applied (3):
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings

Patches absorbed by generator (1):
  - patch-9417fbdd: chore: update SDK version to v2.13.0
    The generator now produces these customizations natively.
Generated by Fern
CLI Version: unknown
Generators:
  - fernapi/fern-python-sdk: 4.37.0
Patches applied (3):
  - patch-13d8e068: chore: ignore local venv and planning docs
  - patch-a065088b: fix(vendors): send Speechmatics STT key instead of api_key
  - patch-33a436b3: fix: address v2.8.0 release review findings
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant