A sophisticated integration skill for building emotionally aware, voice-based AI agents using Hume EVI, LangGraph state management, and Twilio.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install hume-evi-langgraph
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install hume-evi-langgraph using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Hume EVI + LangGraph Integration?
The Hume EVI + LangGraph Integration is a specialized technical framework designed for developers building the next generation of voice-driven AI agents. By combining the emotional intelligence of Hume's Empathic Voice Interface (EVI) with the robust state management of LangGraph, this skill enables the creation of complex, multi-stage conversational workflows that can handle real-world interruptions and long-running call lifecycles.
This implementation focuses on solving the critical challenge of maintaining agent state across asynchronous events. It utilizes the LangGraph interrupt/resume pattern to pause execution during an active call and resume automatically once the call ends, allowing for post-call analysis, coaching, and data storage. It is an essential component for any developer utilizing Openclaw Skills to build high-fidelity voice applications.
Hume EVI + LangGraph Integration Use Cases
- Creating emotionally aware AI customer support agents that react to caller frustration or satisfaction.
- Building automated voice coaching systems that provide feedback based on a user's emotional arc during a conversation.
- Implementing secure voice verification systems that transition from PIN entry to natural language dialogue.
- Developing automated post-call analysis pipelines that extract detailed emotion timelines and transcripts for CRM storage.
How Hume EVI + LangGraph Integration Works
- The process begins by receiving a call via Twilio and optionally verifying a user PIN within the LangGraph StateGraph.
- A dynamic Hume EVI configuration is generated, specifying personas, voice models (like KORA or ITO), and lower temperature settings to ensure professional responses.
- The agent generates a TwiML redirect, handing the call control over to Hume's EVI interface while embedding the necessary API keys and config IDs.
- The LangGraph workflow enters an interrupt state, waiting for a webhook trigger while the human and AI converse.
- Upon call completion, a Hume webhook triggers the resumption of the LangGraph thread using the specific chat session ID.
- The skill fetches the transcript and emotion data via the chat_group endpoint, bypassing known API limitations.
- Final analysis and emotion extraction are performed before the state is saved and the workflow completes.
Hume EVI + LangGraph Integration Setup
To implement this skill, ensure you have your Hume AI and Twilio credentials configured. Install the necessary dependencies:
pip install langgraph httpx twilio
Configure your webhook endpoint to handle the chat_ended event from Hume and map the config_id to your LangGraph thread_id to ensure session continuity.
Hume EVI + LangGraph Integration Data Schema & Taxonomy
The skill manages a complex state object throughout the call lifecycle. Data is organized as follows:
| Field | Type | Description |
|---|---|---|
thread_id |
String | Unique identifier for the LangGraph persistence layer. |
chat_id |
String | The specific Hume EVI session identifier. |
emotion_timeline |
List | A collection of user turns including text and top-5 detected emotions. |
hume_config |
Object | The dynamic JSON configuration sent to Hume API. |
transcript |
List | The full sequence of message objects including message_text and emotion_features. |
Hume EVI + LangGraph Integration Advanced Features
- State Management Interrupts: Sophisticated use of the interrupt/resume pattern to handle asynchronous webhook-driven workflows.
- Emotion Timeline Extraction: Advanced filtering of the Hume emotion feature set (48+ emotions) to identify the top 5 emotional drivers per turn.
- Dynamic Persona Switching: Capability to generate unique EVI configurations on-the-fly based on user input or database lookups.
- Robust Error Mitigation: Specific patterns to avoid common Hume API 404 errors by utilizing chat_group event fetching.
SKILL.md
Loading
Related Openclaw Skills
Cron Expression Helper
A comprehensive tool for creating, validating, and explaining cron expressions through interactive CLI scripts.
Machine Config Migrator
A low-risk, component-based migration tool for transferring development environments and productivity configurations across workstations.
Supabase Dashboard Builder
A lightweight toolkit for building framework-free admin dashboards and data visualizations powered by the Supabase REST API.
PDF Extractor Skill
A specialized tool for extracting text and complex LaTeX mathematical formulas from academic papers in English and Chinese.
PaySpawn
PaySpawn provides AI agents with secure on-chain spending limits and credentials on the Base network without requiring raw private key access.
EvoMap High-Productivity Skill Pack
A robust collection of utility skills designed to enhance AI agent resilience, infrastructure stability, and communication reliability.
Featured*
