VoiceAgentRAG: Solving the RAG Latency Bottleneck in Real-Time Voice Agents Using Dual-Agent Architectures
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908861123788800 |
|---|---|
| author | Qiu, Jielin Zhang, Jianguo Chen, Zixiang Yang, Liangwei Zhu, Ming Tan, Juntao Chen, Haolin Zhao, Wenting Murthy, Rithesh Ram, Roshan Prabhakar, Akshara Heinecke, Shelby Xiong, Caiming Savarese, Silvio Wang, Huan |
| author_facet | Qiu, Jielin Zhang, Jianguo Chen, Zixiang Yang, Liangwei Zhu, Ming Tan, Juntao Chen, Haolin Zhao, Wenting Murthy, Rithesh Ram, Roshan Prabhakar, Akshara Heinecke, Shelby Xiong, Caiming Savarese, Silvio Wang, Huan |
| contents | We present VoiceAgentRAG, an open-source dual-agent memory router that decouples retrieval from response generation. A background Slow Thinker agent continuously monitors the conversation stream, predicts likely follow-up topics using an LLM, and pre-fetches relevant document chunks into a FAISS-backed semantic cache. A foreground Fast Talker agent reads only from this sub-millisecond cache, bypassing the vector database entirely on cache hits. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_02206 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | VoiceAgentRAG: Solving the RAG Latency Bottleneck in Real-Time Voice Agents Using Dual-Agent Architectures Qiu, Jielin Zhang, Jianguo Chen, Zixiang Yang, Liangwei Zhu, Ming Tan, Juntao Chen, Haolin Zhao, Wenting Murthy, Rithesh Ram, Roshan Prabhakar, Akshara Heinecke, Shelby Xiong, Caiming Savarese, Silvio Wang, Huan Sound We present VoiceAgentRAG, an open-source dual-agent memory router that decouples retrieval from response generation. A background Slow Thinker agent continuously monitors the conversation stream, predicts likely follow-up topics using an LLM, and pre-fetches relevant document chunks into a FAISS-backed semantic cache. A foreground Fast Talker agent reads only from this sub-millisecond cache, bypassing the vector database entirely on cache hits. |
| title | VoiceAgentRAG: Solving the RAG Latency Bottleneck in Real-Time Voice Agents Using Dual-Agent Architectures |
| topic | Sound |
| url | https://arxiv.org/abs/2603.02206 |