can tools be called in parallel?
# ask-ai
b
can tools be called in parallel?
t
Thanks for reaching out, Aleh. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Thanks — I’m responding about whether tools can be called in parallel. Yes — LiveKit Agents support calling tools in parallel. This is controlled by the
parallel_tool_calls
parameter available across multiple LLM plugins. When
parallel_tool_calls
is set to
true
, the model can invoke multiple tools simultaneously, which can improve performance for workflows that need several independent calls. A few quick notes: • The parameter is optional and may be disabled by default depending on the provider or plugin you’re using. • If you want, tell me which plugin or provider you’re using (for example: Fireworks, DeepSeek, Perplexity, Azure OpenAI, Anthropic, etc.), and I’ll give plugin-specific guidance or an example of how to enable
parallel_tool_calls
. Sources: Fireworks AI LLM plugin guide | LiveKit Documentation | DeepSeek LLM plugin guide | LiveKit Documentation | Together AI LLM plugin guide | LiveKit Documentation | Perplexity LLM plugin guide | LiveKit Documentation | Telnyx LLM plugin guide | LiveKit Documentation
b
same for STS models?
t
Thanks — I’m responding to whether the same parallel tool-calling capability applies to STS models. I do not have information that confirms whether
parallel_tool_calls
is supported for STS models. The documentation and plugin notes describe
parallel_tool_calls
for LLM plugins (OpenAI, Anthropic, etc.), but they do not mention STS models or whether the parameter applies to them. Could you tell me which STS model or provider you mean (name/version and where you’re configuring it)? With that, I can check more precisely and either show how to enable parallel calls if it’s supported or suggest alternative approaches. Sources: OpenAI LLM plugin guide | LiveKit Documentation
b
gpt-realtime
t
Thanks — you’re asking whether the
gpt-realtime
model supports parallel tool calls. Yes. LiveKit Agents supports parallel tool calls for
gpt-realtime
via the OpenAI Realtime integration. You can enable this by setting
parallel_tool_calls=True
(or the equivalent) on the
RealtimeModel
/
MultimodalAgent
configuration. Enabling it lets the model invoke multiple tools simultaneously, which can speed up workflows that require several independent tool calls. If you’d like, tell me: • How you’re instantiating the model (code snippet or config), and • Which tools you plan to call in parallel. I can then give a concrete example of the configuration or point out anything to watch for (race conditions, ordering needs, or shared-resource considerations). Sources: Azure OpenAI LLM plugin guide | LiveKit Documentation | OpenAI and LiveKit partner to turn Advanced Voice into an API