apiVersion: capsule.dev/v0.1
kind: Capsule
name: cverse-voice-llm-inference-adapter
version: 0.1.0
type: adapter
purpose:
summary: 'Provides adapters for voice-enabled Large Language Models, including specific
integrations for Doubao and Qwen Omni, and a Persona Agent framework for complex
conversational flows. This allows the core inference service to handle real-time
voice interactions.
'
owns:
- Base VoiceLLM plugin interface
- Doubao VoiceLLM integration (protocol, config, real-time client)
- Qwen Omni real-time integration
- Persona Agent framework (supervisor, subagents, tools, schemas)
does_not_own:
- The core inference service
- Other AI model types (ASR, LLM, TTS, Avatar, etc.)
interfaces:
provides:
- kind: library
name: voice-llm-plugin
description: VoiceLLM implementations conforming to the inference core's plugin
interface.
requires:
- kind: library
name: llm-plugin
from_capsule: cverse-llm-inference-adapter
description: Uses LLM plugins for text-based reasoning within the Persona Agent.
dependencies:
capsules:
- name: cverse-inference-core
version: '>=0.1.0'
- name: cverse-llm-inference-adapter
version: '>=0.1.0'
agent:
summary_for_ai: 'An AI agent working on this capsule would focus on integrating
new real-time conversational AI models, enhancing the Persona Agent''s capabilities
(e.g., adding new tools, improving agent reasoning), or optimizing real-time interaction
flows. It requires knowledge of conversational AI, agent frameworks, and Python
plugin development.
'
avoid:
- Modifying the core inference service or other AI model types.
verification:
invariants:
- VoiceLLM plugins must process real-time audio/text inputs and generate appropriate
responses.
- Persona Agent must maintain conversational context and execute tasks reliably.
- Plugins must adhere to the defined VoiceLLM plugin interface.
x-reuse:
notes: 'The `doubao_config.py` contains specific Doubao API configurations. The
Persona Agent''s default tools and agent logic might need customization for different
use cases or domains.
'
x-reconstruct:
install: install.json