Release Notes: v0.10.0¶

Release Date: 2026-06-15

Overview¶

OmniAgent v0.10.0 brings full-duplex voice gateway support, semantic memory capabilities, and enhanced profile modes. This release focuses on real-time voice interactions and long-term memory for conversational AI.

Highlights¶

Voice Gateway - Full-duplex phone calls via Twilio and Telnyx with native voice-to-voice support
Semantic Memory - Long-term memory with vector retrieval via omniretrieve
Auto-Reply Mode - Profile mode for automated responses without human intervention
Commentary Mode - Profile mode for passive observation with optional commentary
RCS Messaging - Rich Communication Services support via Twilio Messaging Service SID
Tools Accessor - Runtime access to registered tools via agent.Tools()
OpenRouter Provider - Additional LLM provider for model routing
Transitive Dependencies - Voice providers are now indirect dependencies via omnivoice

New Features¶

Voice Gateway¶

Full-duplex phone calls with support for both native voice-to-voice and traditional pipeline modes:

# Native voice-to-voice (lowest latency, ~100ms)
omniagent voice serve \
  --listen :8081 \
  --public-url https://your-server.com \
  --gateway twilio \
  --realtime openai \
  --realtime-voice alloy

# Traditional pipeline (custom STT/TTS)
omniagent voice serve \
  --listen :8081 \
  --public-url https://your-server.com \
  --gateway telnyx \
  --stt deepgram \
  --tts elevenlabs \
  --llm anthropic

Gateway Providers¶

Provider	Features
Twilio	Media Streams, native voice-to-voice (OpenAI/Gemini), outbound calls
Telnyx	Media Streaming, traditional pipeline, outbound calls

Voice Modes¶

Mode	Latency	Description
`realtime`	~100-200ms	Native voice-to-voice via OpenAI Realtime or Gemini Live
`pipeline`	~500-1500ms	Traditional STT → LLM → TTS pipeline

Semantic Memory Skill¶

Long-term memory with semantic retrieval for context-aware conversations:

import "github.com/plexusone/omniagent/skills/compiled/memory"

memorySkill, err := memory.New(memory.Config{
    Retriever: retriever, // omniretrieve instance
    TopK:      5,
})

a, err := agent.New(config, agent.WithCompiledSkill(memorySkill))

The memory skill provides these tools:

Tool	Description
`memory_store`	Store information with semantic embedding
`memory_recall`	Retrieve relevant memories by query
`memory_forget`	Remove specific memories

Profile Modes¶

New profile modes for different interaction styles:

import "github.com/plexusone/omniagent/agent/profiles"

// Auto-reply mode: agent responds automatically
profile := &profiles.BootstrapProfile{
    Name: "auto-responder",
    Mode: profiles.ModeAutoReply,
}

// Commentary mode: passive observation with optional input
profile := &profiles.BootstrapProfile{
    Name: "observer",
    Mode: profiles.ModeCommentary,
}

Mode	Behavior
`ModeDefault`	Standard interactive mode
`ModeAutoReply`	Automatic responses without confirmation
`ModeCommentary`	Passive observation, comments when relevant

RCS Messaging¶

Rich Communication Services support via Twilio Messaging Service SID:

channels:
  twilio_sms:
    enabled: true
    account_sid: ${TWILIO_ACCOUNT_SID}
    auth_token: ${TWILIO_AUTH_TOKEN}
    phone_number: ${TWILIO_PHONE_NUMBER}
    messaging_service_sid: ${TWILIO_MESSAGING_SERVICE_SID}  # Enables RCS

With Messaging Service SID configured:

RCS messages are sent when supported by the recipient
Automatic fallback to SMS/MMS when RCS unavailable
Rich media, read receipts, and typing indicators

Tools Accessor¶

Runtime access to registered tools:

a, err := agent.New(config, agent.WithCompiledSkill(mySkill))

// Access tools at runtime
tools := a.Tools()
for _, tool := range tools {
    fmt.Printf("Tool: %s - %s\n", tool.Name, tool.Description)
}

OpenRouter Provider¶

Additional LLM provider for model routing:

agent:
  provider: openrouter
  model: anthropic/claude-sonnet-4
  api_key: ${OPENROUTER_API_KEY}

New CLI Commands¶

Voice Commands¶

Command	Description
`omniagent voice serve`	Start the voice gateway server
`omniagent voice status`	Show voice configuration status
`omniagent voice call NUM`	Make an outbound call

Voice Serve Flags¶

Flag	Description
`--listen`	Listen address (default `:8081`)
`--public-url`	Public URL for webhooks
`--gateway`	Gateway provider: `twilio`, `telnyx`
`--realtime`	Realtime provider: `openai`, `gemini`
`--realtime-voice`	Voice for realtime API
`--stt`	STT provider for pipeline mode
`--tts`	TTS provider for pipeline mode
`--llm`	LLM provider for pipeline mode
`--ngrok`	Enable ngrok tunnel
`--ngrok-domain`	Custom ngrok domain

Configuration Changes¶

New Fields¶

Field	Type	Description
`channels.twilio_sms.messaging_service_sid`	string	Messaging Service SID for RCS
`voice.gateway`	string	Gateway provider: `twilio`, `telnyx`
`voice.realtime.provider`	string	Realtime provider: `openai`, `gemini`
`voice.realtime.voice`	string	Voice for realtime API

Voice Gateway Configuration¶

voice:
  enabled: true
  gateway: twilio  # or: telnyx

  # Native voice-to-voice (recommended for low latency)
  realtime:
    provider: openai   # or: gemini
    voice: alloy       # OpenAI: alloy, nova, etc.

  # Traditional pipeline (for custom STT/TTS)
  # stt:
  #   provider: deepgram
  # tts:
  #   provider: elevenlabs
  # llm:
  #   provider: anthropic
  #   model: claude-sonnet-4-20250514

  # Twilio credentials
  twilio:
    account_sid: ${TWILIO_ACCOUNT_SID}
    auth_token: ${TWILIO_AUTH_TOKEN}
    phone_number: ${TWILIO_PHONE_NUMBER}

  # Telnyx credentials
  # telnyx:
  #   api_key: ${TELNYX_API_KEY}
  #   connection_id: ${TELNYX_CONNECTION_ID}
  #   phone_number: ${TELNYX_PHONE_NUMBER}

Architecture Changes¶

Transitive Dependencies¶

Voice provider packages are now indirect dependencies via omnivoice:

Before v0.10.0:

import (
    "github.com/plexusone/omnivoice"
    "github.com/plexusone/omni-openai/omnivoice/realtime"
    "github.com/plexusone/omni-twilio/omnivoice/gateway"
)

After v0.10.0:

import (
    "github.com/plexusone/omnivoice"
    _ "github.com/plexusone/omnivoice/providers/all"
)

// Get realtime factory by name
factory, _ := omnivoice.GetRealtimeFactory("openai")

// Configure gateway tools using generic types
tools := []omnivoice.ToolDefinition{...}
opts := []registry.ProviderOption{
    omnivoice.WithGatewayTools("twilio", tools),
}

This reduces go.mod complexity and ensures provider versions stay in sync.

Dependency Updates¶

Package	From	To
`omnivoice`	v0.10.0	v0.11.0
`omnillm`	v0.15.5	v0.16.0
`omnichat`	v0.7.0	v0.8.0
`wazero`	v1.11.0	v1.12.0
`kin-openapi`	v0.139.0	v0.140.0
`grokify/mogo`	v0.74.5	v0.74.6

New Transitive Dependencies¶

These packages are now indirect (via omnivoice):

Package	Purpose
`omni-google`	Gemini Live realtime provider
`omni-openai`	OpenAI Realtime provider
`omni-telnyx`	Telnyx gateway provider
`omni-twilio`	Twilio gateway provider

Upgrade Guide¶

From v0.9.0¶

Update your dependency:

go get github.com/plexusone/omniagent@v0.10.0
go mod tidy

(Optional) Add voice gateway support:

voice:
  enabled: true
  gateway: twilio
  realtime:
    provider: openai
    voice: alloy
  twilio:
    account_sid: ${TWILIO_ACCOUNT_SID}
    auth_token: ${TWILIO_AUTH_TOKEN}
    phone_number: ${TWILIO_PHONE_NUMBER}

(Optional) Add semantic memory:

import "github.com/plexusone/omniagent/skills/compiled/memory"

memorySkill, _ := memory.New(memory.Config{Retriever: retriever})
a, _ := agent.New(config, agent.WithCompiledSkill(memorySkill))

(Optional) Enable RCS messaging:

channels:
  twilio_sms:
    messaging_service_sid: ${TWILIO_MESSAGING_SERVICE_SID}

Breaking Changes¶

None. This release is fully backward compatible with v0.9.0.

Documentation¶

New guides added:

Voice Integration - Voice gateway setup and configuration
Memory Skill - Semantic memory with vector retrieval

Updated guides:

Twilio Setup - RCS and Messaging Service SID
LLM Providers - OpenRouter provider

Full Changelog¶

See CHANGELOG.md for the complete list of changes.