Deployment Prerequisites
1. System Requirements
| Item | Requirement | Notes |
|---|---|---|
| OS | CentOS 8+ / RHEL 8+ / Ubuntu 22.04+ | For Kylin / UOS, evaluate against the corresponding RHEL/Ubuntu baseline |
| CPU architecture | x86_64 (amd64) / arm64 | |
| Kernel | Linux 5.13+ | uname -r |
| CPU | 8 cores or more | |
| Memory | 16 GB or more | |
| Disk | 200 GB+ free space | Data disk SSD recommended |
2. Software Dependencies
| Software | Version | Verify command |
|---|---|---|
| Docker Engine | 24.x or later | docker --version |
| Docker Compose | v2.x (not v1) | docker compose version |
| Base tools | tar, openssl | tar --version && openssl version |
| cgroup v2 | Enabled (unified hierarchy) | stat -fc %T /sys/fs/cgroup/ should output cgroup2fs, or `mount |
3. Offline Package
| Item | Detail |
|---|---|
| Filename | chatkit-offline-release-1.0.xx-linux-amd64-xxxxxxxx.tar.gz |
| Size | ~ 4 GB |
| Contents | All microservice images + AI runtime (ai-infra-rs) + deployment scripts |
| Storage location | Data disk (e.g. /data/), not the system disk / |
4. LLM Model Information (Admin Console)
FinClaw relies on external LLM providers for inference. Obtain the following key information from your model provider in advance, and configure it in the admin console (Admin) after deployment:
| Item | Description | Example |
|---|---|---|
| Provider | LLM provider (openai / anthropic / deepseek / qwen / ollama / litellm / custom, etc. supported) | deepseek |
| Model name | Model identifier | deepseek-v4-flash |
| API Key | API key | sk-xxxxxxxx... |
| Base URL | Custom service URL (if needed) | https://api.deepseek.com/v1 |
5. Port Requirements
FinClaw uses ports in the 261xx ~ 264xx range by default, with the API Gateway (26100) as the external entrypoint.
| Purpose | Port | External exposure required |
|---|---|---|
| API Gateway (unified entrypoint) | 26100 | ✅ Required |
| Other services | 261xx – 265xx | ❌ Internal |
See Deployment Components for the full port-to-service mapping.