v6.1.1

6.1.1 (2026-03-27)

Bug Fixes

• websockets: restore optional message param on control send_ methods (#680) (0018fc4)

v6.1.0

6.1.0 (2026-03-26)

Features

• agent: support multi-provider speak/think configuration and typed listen parameters (#676) (5dfb1aa)

v6.0.1

6.0.1 (2026-02-24)

Bug Fixes

• 🌿 skip_validation:true to allow unknown messages back from the API (#669) (48354d2)

v6.0.0

6.0.0 (2026-02-23)

⚠ BREAKING CHANGES

• promote v6.0.0-rc.2 to v6.0.0 stable
• Fully generated WebSocket clients for Listen v1/v2, Speak v1, and Agent v1, replacing all hand-rolled WebSocket code from v5
• Simplified send methods: send_media() accepts raw bytes directly, control messages use dedicated methods (send_keep_alive(), send_finalize(), send_flush()) instead of generic send_control()
• New domain-specific type system (deepgram.listen.v1.types, deepgram.agent.v1.types) replacing shared barrel imports from deepgram.extensions.types.sockets
• Agent settings types renamed to match generated schema hierarchy (e.g. AgentV1SettingsMessage → AgentV1Settings)

Features

• promote v6.0.0-rc.2 to v6.0.0 stable (34f543e)
• feat!: v6.0.0 stable — fully generated SDK with WebSocket support
• Custom WebSocket transport support for alternative runtimes
• TextBuilder class for TTS pronunciation and pause controls
• SageMaker transport extracted to separate deepgram-sagemaker package
• 22 production-ready examples covering auth, transcription, voice agents, TTS, text intelligence, and management APIs
• CI matrix testing across Python 3.8–3.13

v6.0.0-rc.2

6.0.0-rc.2 (2026-02-18)

Bug Fixes

• sagemaker: extract SageMaker transport to separate deepgram-sagemaker package (#665) (e6317c5)

Refactors

• sagemaker: extract SageMaker transport to separate package (#663) (d82b699)
• sagemaker: move SageMaker transport to separate package (#662) (16d500e)

v6.0.0-rc.1

⚠ BREAKING CHANGES

• v6 — fully generated SDK with latest APIs and WebSocket support (#640)

Features

• helpers: add TextBuilder class for TTS pronunciation and pause controls (#660)
• sagemaker: add SageMaker transport for running Deepgram on AWS SageMaker endpoints (#659)
• v6 — fully generated SDK with latest APIs and WebSocket support (#640)
• websockets: add custom WebSocket transport support (#658)

Install

pip install deepgram-sdk==6.0.0rc1

v5.3.2

5.3.2 (2026-01-29)

Bug Fixes

• speak: correct TTS warning event field names to match API response (#653) (f7ab1da)
• types: change speaker and related fields from float to int (#652) (00ee485), closes #641
• websockets: support array parameters in Listen v1 and v2 clients (#650) (38cc1e5)

Miscellaneous Chores

• release 5.3.2 (d61ce8c)

v6.0.0-beta.2

Release Notes: v6.0.0-beta.2

Overview

This beta release (v6.0.0-beta.2) includes critical fixes for binary audio handling across all WebSocket clients, SDK regeneration with the latest API specifications, and important improvements to the Agent, Listen, and Speak modules.

🎯 What's Changed

🐛 Bug Fixes

Binary Audio Support Improvements

• Listen (WebSocket): Fixed send_media() parameter type from str to bytes for proper binary audio handling in both V1 and V2 clients
• Speak (WebSocket): Added comprehensive binary audio response support in socket client
  • Updated response union type from str to bytes
  • Enhanced recv(), __iter__(), __aiter__(), and start_listening() methods to handle both binary bytes and JSON text messages
• Agent (WebSocket):
  • Added missing send_media(message: bytes) method to both async and sync Agent V1 clients
  • Updated response union type from str to bytes for binary audio data
  • Enhanced all response handling methods to support both binary bytes and JSON text messages

🔄 SDK Updates

• SDK Regeneration: Complete regeneration of SDK with latest API specifications
  • Added entity detection support in Listen V1 with new entities field in transcription results
  • Updated Agent V1 think provider models with support for:
    • Anthropic models
    • AWS Bedrock with new credential handling
    • Google models
    • Groq models
  • Improved Agent V1 speak provider structure with dedicated provider classes (Cartesia, Deepgram, ElevenLabs, OpenAI)
  • Enhanced Agent V1 listen provider with V1 and V2 configurations
  • Added comprehensive HTTP client improvements and test coverage

📚 Documentation & Configuration

• Updated .fernignore to protect manual socket client fixes from regeneration
• Added .github, docs, and examples folders to fernignore
• Documented all manual binary audio fixes in socket clients

📊 Stats

• 107 files changed: 1,940 insertions(+), 1,036 deletions(-)
• Key modules affected: Agent, Listen, Speak, Core HTTP Client

⚠️ Breaking Changes

• send_media() methods now require bytes instead of str for all WebSocket clients (Listen V1, Listen V2, Agent V1, Speak V1)

📦 Installation

pip install deepgram-sdk==6.0.0b2

---

Full Changelog: v6.0.0-alpha.4...v6.0.0-beta.2

v6.0.0-alpha.4

Full Changelog: v5.3.0...v6.0.0-alpha.4

v5.3.0

5.3.0 (2025-10-30)

Features

• add projects billing fields list methods (#621) (10d67cd)