Function calling

Function calling lets an LLM invoke approved tools or APIs during a conversation, so an avatar can fetch data, perform actions, or answer from live systems instead of relying on text generation alone.

How function calling works

Function calling lets a language model ask an external tool or API for help during a conversation. Instead of only generating text, the model can trigger a defined action such as checking an order, searching a knowledge base or booking a meeting.

In a real-time avatar system, function calling sits inside the agent loop. The avatar hears the user, the model decides whether it needs a tool, the tool returns structured data, and the avatar turns the result into a spoken response.

A concrete example: a customer asks whether their plan includes a feature. The agent calls an account API, reads the plan data, and answers through the avatar without asking the user to leave the conversation.

The key is control. Functions should be narrow, typed and permissioned so the avatar can take useful actions without inventing data or performing something the application did not allow.

What Anam ships

Anam's Cara-4 model delivers expressive real-time avatars with around 150 ms server-side avatar-generation latency once a session is running, across 70+ languages. Builders use JavaScript and Python SDKs or integrations for LiveKit, Pipecat, ElevenLabs Agents, Agora, and VideoSDK. Bring any AI stack including OpenAI, Claude, Gemini, Mistral, Groq, Deepgram, Cartesia, or custom providers. The platform supports WebRTC delivery, SOC 2 Type II, HIPAA, zero data retention, and regional data residency. Sessions stream low-latency audio and video to browsers and native apps.

Explore the real-time avatar API →

Related terms

RAG

RAG, or retrieval-augmented generation, is a pattern where an agent retrieves relevant documents or data at query time and uses them to produce a more accurate, grounded response.

Grounding

Grounding is the practice of anchoring a model's answer to trusted source content, retrieved documents, or tool results so an avatar can respond with verified information instead of guessing.

Prompt

A prompt is the instruction and context given to a language model for a specific response, including the user's message, system rules, retrieved content, and tool results when available.

AI sales agent

An AI sales agent is an interactive avatar or voice agent configured to qualify prospects, answer product questions, guide buying decisions, and hand over to a human or booking flow when a lead is ready.

Frequently asked questions

What does function calling let an avatar agent do?

Function calling lets an avatar agent take structured actions, such as checking account data, booking a meeting, searching a knowledge base, or triggering a workflow during the conversation.

How is function calling different from prompting?

Prompting shapes what the model says. Function calling gives the model a controlled way to request a specific tool or action with structured inputs that software can execute safely.

Why does function calling matter for real-time avatars?

It turns the avatar from a talking interface into a useful agent. The avatar can answer from live systems, complete tasks, and keep the user in a natural conversation while work happens behind the scenes.

What can go wrong with function calling?

Poor tool design can make the agent slow, unsafe, or unpredictable. Each function needs clear permissions, validation, fallback behavior, and latency that fits the live conversation.

Last updated: 17th July 2026 · Reviewed quarterly.

← Back to glossary