Developer digest

Write the Sonora Developer Digest, Sonora's email to developers building on Sonora Ensemble, the voice-agent platform. Recipients are developers who opted in to the digest. This issue is dated April 10, 2026 and can be viewed in the browser at https://sonora-email.s3.amazonaws.com/digest/2026-04-10

Digest summary: this issue covers telephony voice agents, reusable agent configurations, JavaScript and Python SDK releases, Larkspur LLM support, and Dictate 3 Chinese.

Five product updates, in this order:

• Reusable Agent Configurations for the Ensemble API. Agent configurations are stored and managed through the Ensemble API, then referenced by ID instead of sending the full configuration on every WebSocket session. Teams can A/B test voices and prompts, manage per-customer agents, and support multi-agent architectures without a code deploy. Docs: https://sonora-email.s3.amazonaws.com/docs/agent-configurations
• JavaScript SDK v4 and Python SDK v3 are now GA. Both releases are stable and production-ready, with updated APIs, improved type safety, and streamlined authentication. Migration guides list breaking changes and upgrade paths. Blog: https://sonora-email.s3.amazonaws.com/blog/sdk-ga
• Inbound and outbound telephony voice agents. Two production-patterned reference implementations for telephony agents on the Ensemble API. Inbound (https://sonora-email.s3.amazonaws.com/docs/telephony/inbound) handles calls with function calling and barge-in. Outbound (https://sonora-email.s3.amazonaws.com/docs/telephony/outbound) adds answering machine detection and voicemail. Both deploy in one command. Docs: https://sonora-email.s3.amazonaws.com/docs/telephony
• Larkspur is now available as an Ensemble LLM provider. Larkspur Reason 49B handles multi-step agent reasoning and Larkspur Nano 8B delivers cost-efficient performance for targeted tasks. Both are in the Standard pricing tier. Set the provider type to larkspur. Docs: https://sonora-email.s3.amazonaws.com/docs/llm-providers/larkspur
• Dictate 3 adds Chinese Mandarin, Simplified and Traditional. Dictate 3 supports Simplified Chinese (zh, zh-CN, zh-Hans) and Traditional Chinese (zh-TW, zh-Hant). Setting the model to dictate-3 with the matching language code transcribes Mandarin audio in both streaming and batch requests. Playground: https://sonora-email.s3.amazonaws.com/playground

Each update closes with a text link; the only button in the email is the webinar Register button.

Webinar: AI-Powered Outbound Dialing in Healthcare. Sonora and Cobalt Cloud show how voice agents handle clinical trial screening calls with real conversational flow, with two live reference architectures: Sonora speech models on Cobalt's managed model platform, and Cobalt's contact center suite for a fully managed path. When: Thu, Apr 16 · 9 AM PT / 12 PM ET. Speakers: Ines Marlow, Director of Product at Sonora, and Tomas Reyes, Partner Solutions Architect at Cobalt Cloud. Register: https://sonora-email.s3.amazonaws.com/events/healthcare-outbound-dialing

Quick Hits, six links with a one-line description each:

• The Definitive Guide to Voice AI Agents [E-book]: architecture-level playbook covering the full voice agent stack, four build-approach trade-offs, conversational UX design, performance diagnostics, and compliance architecture. https://sonora-email.s3.amazonaws.com/guides/voice-agents-ebook
• Low Latency Voice AI: What It Is and How to Achieve It: how to hit sub-300ms voice AI latency using streaming recognition, real-time LLM processing, and enterprise deployment strategies. https://sonora-email.s3.amazonaws.com/blog/low-latency-voice-ai
• Which STT API Handles Production Reality?: real-world accuracy, latency, pricing, and production scalability compared for enterprise speech-to-text. https://sonora-email.s3.amazonaws.com/blog/stt-production-comparison
• WebSocket vs REST for Text-to-Speech: When to Use Which: a decision framework for choosing between WebSocket and REST TTS APIs based on latency requirements, telephony, and voice agent use cases. https://sonora-email.s3.amazonaws.com/blog/websocket-vs-rest-tts
• Barge-In, Interruptions, and Turn-Taking: when interruption handling holds up in call center deployments, and when noisy audio and concurrency demand custom speech infrastructure. https://sonora-email.s3.amazonaws.com/blog/barge-in-and-turn-taking
• How AI Contact Centers Detect Caller Intent: how speech recognition, NLU, and classification work together for intent detection, with accuracy benchmarks and latency requirements. https://sonora-email.s3.amazonaws.com/blog/caller-intent-detection

Reference artwork: assets.illustrations.digestBanner for the digest masthead, assets.illustrations.agentConfigs for the agent configurations update, assets.illustrations.webinarHealthcare for the webinar.

The footer gives the postal address and links to https://sonora-email.s3.amazonaws.com/unsubscribe and https://sonora-email.s3.amazonaws.com/preferences

Dark mode should be supported.

Use /world/sonora-email for brand chrome and CSS. Use the absolute HTTPS URLs from tokens.json; do not use /world/ paths as image src.

Leave the HTML in output/email.html, a plain-text version in output/email.txt, and the subject in output/meta.json as {"subject": "..."}. Don't include From, To, or Date headers.