基于 FreeSWITCH、Spring Boot 和 Vue 3 的开源呼叫中心系统，支持多租户、人工呼入呼出、智能外呼任务、ACD 技能组路由、IVR、WebRTC 坐席与录音管理，适用于企业客服和外呼业务，也可作为 AI 语音应用的呼叫底座。

An enterprise-grade AI call center system built on FreeSwitch and Spring Boot, providing multi-tenant, multi-skill-group, and intelligent routing — a complete set of call center capabilities.

中文 | English

🌟 If this project helps you, please give it a ⭐ Star! 🙏

Official site: https://dongjb741280.github.io/ai-call-center/

This is an AI call center solution built on a microservices architecture, integrating FreeSwitch as the call engine and providing complete call management, agent management, intelligent routing, monitoring and statistics. It supports multi-company and multi-skill-group management, with enterprise-grade high-concurrency and high-availability characteristics.

This project is primarily intended for large-scale human inbound and outbound calling scenarios, or as the foundation for AI-powered unattended calling.

For a pure AI call center (unattended-first), please refer to:

https://github.com/dongjb741280/IntelliCall-Pro

Dashboard	Agent Management

IVR Management	Agent Monitor

flowchart LR
    AG[Agent Console<br/>Vue3 + JsSIP Softphone] -- WebSocket --> CALL
    CP[Caller / Customer] -- SIP/RTP --> FS[FreeSWITCH<br/>Media & Signaling]
    FS <-- "ESL events / commands" --> CALL[voxai-call<br/>Call Control :7200]
    FS -. Local recording files .-> MEDIA[FS_MEDIA<br/>Recording File Server :7430]
    CALL -- HTTP GET recordings --> MEDIA
    CALL -- HTTP --> IVR[voxai-ivr<br/>IVR Engine :7300]
    IVR -- callback --> CALL
    ADMIN[voxai-admin<br/>Admin API :7100] -- config --> CALL
    CALL --> R[(Redis)]
    CALL --> DB[(MySQL)]
    CALL -- Upload recordings --> M[(MinIO)]
    ADMIN --> DB
    ADMIN --> N[Nacos]
    CALL --> N

FS_MEDIA serves FreeSWITCH local recording files over HTTP on port 7430. After hangup, voxai-call downloads recordings from this server and uploads them to MinIO. See Recording Upload Flow.

ai-call-center/
├── voxai-common/          # Core business module (data layer)
├── voxai-admin/           # Admin API service (port 7100)
├── voxai-call/            # FreeSwitch integration service (port 7200)
├── voxai-ivr/             # IVR engine service (port 7300)
├── voxai-dialer/          # External outbound scheduler (HTTP client, disabled by default)
├── freeswitch/            # FreeSwitch configuration and Docker image
├── portal/                # Product portal (VoxHub)
├── docs/                  # Business docs (scope/flows/data model/engine interfaces)
├── .github/workflows/     # GitHub Actions (portal CI/CD)
├── .gitlab-ci.yml         # GitLab CI/CD pipeline
└── Dockerfile             # Multi-stage Docker build

• Backend framework: Spring Boot 2.3.12
• Microservices: Spring Cloud + Nacos
• Database: MySQL/PostgreSQL/Oracle/SQL Server (multi-database support)
• Cache: Redis
• Message queue: RabbitMQ (optional Kafka)
• File storage: MinIO
• Call engine: FreeSwitch
• Build tool: Maven + Jib (build images without a Docker Daemon)
• Image registry: Harbor
• Containerization: Docker
• CI/CD: GitLab CI + Watchtower auto-deploy

• Entities: company, agent, skill group, call records and other business entities
• Data access: MyBatis mappers
• Utilities: encryption, ID generation and other common utilities

• Administration: agent management, skill group management, company configuration
• Auth: Spring Security + JWT
• Scheduled tasks: Quartz scheduler
• Statistics: call statistics, agent state statistics
• File management: MinIO integration

• Call control: inbound/outbound, transfer, hold, mute, etc.
• Intelligent routing: ACD (Automatic Call Distribution) algorithms
• Agent state: real-time state management and monitoring
• WebSocket: real-time communication
• TCP service: agent state subscription
• Event handling: FreeSwitch event listening and processing

• Flow engine: parses cc_ivr_workflow.content into a node graph, executes by play/input/transfer
• Interaction: playback and digit collection via FreeSwitch playback / play_and_get_digits
• Handoff: calls back voxai-call on completion to route to skill group/agent/VDN/hangup

Independent scheduler with cruise/predictive pacing, command acknowledgements, lease recovery and CDR result submission. See deployment and recovery before enabling.

• Inbound handling: automatic routing to a skill group
• Outbound: agent-initiated calls
• Call transfer: agent-to-agent transfer, skill-group transfer
• Call hold: hold / unhold
• Mute control: agent mute / unmute

• VDN identification: service number matches cc_vdn_phone → cc_vdn_code, selects the active route by cc_vdn_schedule (weekday/date/time range)
• Route types (cc_vdn_config.route_type): skill group / playback / IVR / agent / outbound
• Digit navigation: collect digits with play_and_get_digits after playback, match against cc_vdn_dtmf to route to skill group/IVR/route code/agent extension/hangup
• IVR engine: voxai-ivr parses the cc_ivr_workflow definition, executes play/input/transfer, then hands back

• Multiple assignment strategies:
  • Least answered
  • Least talk time
  • Longest ready
  • Random
  • Round-robin
  • Custom

• Queuing strategies: priority queue, VIP queue
• Overflow handling: queue timeout, skill-group overflow
• Memory agent: match the customer's historical agent

• State management: ready, busy, talking, offline, etc.
• Multiple login methods: SIP, WebRTC, phone
• Access control: role-based permission management

• Real-time monitoring: agent state, call state
• Statistics: call volume, answer rate, average talk time
• Report export: Excel report export

• JDK 1.8+
• Maven 3.6+
• MySQL 5.7+ / PostgreSQL 10+ / Oracle 11g+ / SQL Server 2016+
• Redis 5.0+
• FreeSwitch 1.10+
• Docker & Docker Compose

1. Create the database
2. Initialize the database: table scripts and migration SQL are under docs/sql/ (voxai.sql full database dump, migration_*.sql incremental migrations)
3. Update the database connection settings in the config file

Edit application-dev.properties:

# Database
spring.datasource.url=jdbc:mysql://localhost:3306/voxai
spring.datasource.username=root
spring.datasource.password=your_password

# Redis
spring.redis.host=localhost
spring.redis.port=6379
spring.redis.password=your_redis_password

# MinIO
voxai.minio.endpoint=http://localhost:9000
voxai.minio.access.key=admin
voxai.minio.secret.key=your_secret_key

# Build all modules (skip tests)
mvn clean install -DskipTests

# Build a single module
mvn clean install -pl voxai-admin -DskipTests

# voxai-admin (admin backend, port 7100)
java -jar voxai-admin/target/voxai-admin-1.0.0.jar

# voxai-call (call service, port 7200)
java -jar voxai-call/target/voxai-call-1.0.0.jar

# voxai-ivr (IVR engine, port 7300)
java -jar voxai-ivr/target/voxai-ivr-1.0.0.jar

# Build and push to Harbor
mvn compile jib:build -pl voxai-admin,voxai-call \
  -DsendCredentialsOverHttp=true \
  -Ddocker.repostory=localhost:8444 \
  -Djib.to.tags=latest

# Or build to the local Docker Daemon
mvn compile jib:dockerBuild -pl voxai-admin

# Pull from Harbor and start
docker pull localhost:8444/voxai/voxai-admin:latest
docker run -d --name voxai-admin --network host \
  localhost:8444/voxai/voxai-admin:latest

• voxai-admin: http://localhost:7100/voxai-admin/swagger-ui/index.html
• voxai-call: http://localhost:7200/voxai-call/swagger-ui/index.html

• voxai-admin: admin backend service (port 7100)
• voxai-call: call control service (port 7200)
• voxai-ivr: IVR engine service (port 7300)
• FS_MEDIA: recording file server on the FreeSWITCH media node (HTTP port 7430), used by voxai-call to retrieve local recordings
• freeswitch: call engine (port 5060)

• MySQL: business data storage
• Redis: cache and session storage
• MinIO: recording files and media storage

Compile → Unit Tests → SonarQube → Package → Jib Build (Harbor) → Nexus Publish → 1Panel API Deploy
                                                                               ↓
                                                                    Watchtower (k8s-work-4)
                                                                    Dual safeguard, auto pull & restart

• CI (GitLab): build, test, code scan, Jib pushes images to Harbor
• Harbor: local image registry (localhost:8444)
• Dual CD safeguard:
  • Option A (1Panel API): CI actively triggers images/pull + containers/upgrade
  • Option B (Watchtower): runs in the background on k8s-work-4, periodically checks and auto-deploys
  • Both coexist without conflict; either one succeeding is enough

docker run -d \
  --name watchtower --restart unless-stopped \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v /root/.docker/config.json:/config.json \
  -e DOCKER_CONFIG=/ \
  containrrr/watchtower \
  --interval 30 --cleanup \
  voxai-admin voxai-call

Depends on the Harbor login credentials in /root/.docker/config.json

• Development (dev)
• Production (prod)
• Nacos config center integration

• MySQL (primary)
• PostgreSQL
• Oracle
• SQL Server

• JWT token authentication
• Encrypted password storage
• API permission control

Admin context path /voxai-admin (port 7100), call context path /voxai-call (port 7200).

• POST /voxai-admin/index/login - admin login
• GET /voxai-admin/index/logout - admin logout
• GET /voxai-admin/index/health - admin health check
• GET /voxai-call/index/health - call health check

• GET /voxai-admin/config/agent - agent list
• POST /voxai-admin/config/agent - add agent
• GET /voxai-admin/config/group - skill group list
• POST /voxai-admin/config/group - add skill group
• GET /voxai-admin/config/group/agent/{id} - agents in a skill group

• POST /voxai-call/cti/call/makeCall - make a call
• POST /voxai-call/cti/call/hangup - hang up
• POST /voxai-call/cti/call/hold / cancelHold - hold / unhold
• POST /voxai-call/cti/call/readyMute / cancelMute - mute / unmute
• POST /voxai-call/cti/agent/ready / notReady - set ready / not ready
• POST /voxai-call/cti/agent/logout - agent logout
• GET /voxai-call/cti/admin/call - call list

• GET /voxai-admin/index/health - voxai-admin health check (port 7100)
• GET /voxai-call/index/health - voxai-call health check (port 7200)

• Logback logging configuration
• ELK log collection support
• Structured log output

• Prometheus metrics collection
• JVM performance monitoring
• Database connection pool monitoring

src/main/java/
├── com.voxai.api/          # API service
│   ├── web/                 # Controller layer
│   ├── service/             # Service layer
│   └── configration/        # Configuration classes
├── com.voxai.core/         # Core module
│   ├── entity/              # Entities
│   ├── mapper/              # Data access layer
│   └── enums/               # Enums
└── com.voxai/              # FreeSwitch integration
    ├── cc/                  # Call control
    ├── fs/                  # FreeSwitch event handling
    └── websocket/           # WebSocket communication

• Follow the Alibaba Java Development Manual
• Use a unified exception handling mechanism
• Support multi-tenant data isolation
• Async processing for better performance

A: Refer to the configuration files under freeswitch/, which support both CentOS and Debian.

A: Implement a new assignment algorithm under voxai-call/src/main/java/com/voxai/cc/acd/assign/.

A: Integrate via WebHook callbacks or message queues, supporting HTTP and RabbitMQ.

A: Add new entities and interfaces in the voxai-common module, and implement the concrete business logic in voxai-admin.

This project is licensed under the MIT License. See the LICENSE file for details.

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

• Author: dongjianbin
• Email: [email protected]
• Repository: GitHub Repository

• Integrated Jib image builds, removing the Docker Daemon dependency
• Added GitLab CI + Harbor + Watchtower auto-deploy pipeline
• Unified Harbor environment variable naming (HARBOR_REGISTRY prefix)

• Initial release
• Basic call center features
• FreeSwitch call engine integration
• Multi-tenant management
• Complete API endpoints

Note: This project is for learning and research only. For commercial use, please ensure compliance with relevant laws and regulations.