AI Assistant Setup

Web.Casa includes a built-in AI assistant with 73 tools that can manage your server, diagnose issues, generate configurations, review code, run system health inspections, and auto-heal problems. It natively supports three API protocols, covering six major LLM providers.

Warning: The AI assistant can execute real actions on your server (deploying projects, running commands, managing containers). A confirmation dialog is shown before risky operations, but you should still review each action carefully.

Provider Presets

Web.Casa natively supports 6 providers with 3 API protocols:

Provider	API Format	Base URL	Models
OpenAI	openai-chat	https://api.openai.com	gpt-4o, gpt-4o-mini, o3-mini
Anthropic	anthropic-messages	https://api.anthropic.com	claude-sonnet-4-20250514, claude-haiku-4-20250414
Google Gemini	google-generativeai	https://generativelanguage.googleapis.com	gemini-2.0-flash, gemini-2.5-pro
DeepSeek	openai-chat	https://api.deepseek.com	deepseek-chat, deepseek-reasoner
Moonshot (Kimi)	openai-chat	https://api.moonshot.cn	moonshot-v1-8k, moonshot-v1-32k, moonshot-v1-128k
Zhipu (智谱)	openai-chat	https://open.bigmodel.cn/api/paas	glm-4-flash, glm-4

Any OpenAI-compatible endpoint (Ollama, LM Studio, vLLM) also works by selecting the openai-chat format and entering a custom base URL.

Configuration

1. Navigate to Settings > AI
2. Select a provider preset or choose custom
3. The API format is auto-selected based on the provider
4. Enter the Base URL for the API endpoint
5. Enter your API Key (encrypted with AES-GCM at rest)
6. Select the Model to use
7. Save settings and test the connection

Tip: API keys are encrypted at rest and never exposed in the UI after saving.

Embedding Model (Optional)

For semantic search and vector-based memory retrieval, you can configure a separate embedding model:

Provider	Models
OpenAI	text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002
Zhipu	embedding-3, embedding-2

You can optionally provide separate credentials for the embedding endpoint (Base URL + API Key). If not specified, the main provider’s credentials are used as fallback.

Chat Interface

The AI assistant is accessible through a chat panel in the Web.Casa UI:

• Conversation history — previous messages are maintained within the session
• Fullscreen mode — expand the chat to full screen for complex interactions
• Ctrl+Enter to send messages
• Clear conversation history at any time

Tools (73 Total)

The AI assistant has access to 73 tools organized in 17 categories:

Host Management (7 tools)

• list_hosts — list all reverse proxy hosts
• get_host — get details of a specific host
• create_host — create a new reverse proxy host
• update_host — modify an existing host
• delete_host — remove a host permanently
• toggle_host — enable or disable a host
• clone_host — duplicate a host configuration to a new domain

Project Deployment (10 tools)

• list_projects — list all deployed projects
• get_project — get project details
• create_project — create a new project from a Git repository
• deploy_project — trigger a build and deployment
• get_build_log — retrieve build logs
• get_runtime_log — retrieve runtime logs
• start_project / stop_project — start or stop a project
• rollback_project — rollback to a specific build number
• auto_deploy — one-command: Git URL → project creation → framework detection → build → reverse proxy with HTTPS

Docker Operations (9 tools)

• docker_ps — list all containers with status
• docker_logs — view container logs
• docker_list_stacks — list Compose stacks
• docker_manage_container — start, stop, or restart a container
• docker_run_container — create and run a new container
• docker_pull_image — pull an image from a registry
• docker_get_container_stats — real-time CPU/memory/network/IO stats
• docker_remove_container — remove a container
• docker_prune — clean up unused resources (containers, images, volumes, or all)

Database Operations (5 tools)

• database_list_instances — list MySQL, PostgreSQL, MariaDB, Redis instances
• database_create_instance — create a new database instance in Docker
• database_create_database — create a logical database
• database_create_user — create a database user with access grants
• database_execute_query — execute read-only SQL queries (SELECT, SHOW, DESCRIBE, EXPLAIN)

File Management (5 tools)

• read_file — read file contents (restricted to safe directories)
• list_dir — list directory contents with metadata
• write_file — create or overwrite a file (max 1MB)
• delete_file — delete a file or directory
• rename_file — rename or move a file

System & Metrics (3 tools)

• get_system_metrics — current CPU, memory, disk usage
• get_system_info — hostname, OS, kernel, uptime, CPU cores, architecture
• run_command — execute a shell command (dangerous commands are blocked)

AI-Powered Diagnostics (6 tools)

• suggest_env_vars — recommend environment variables for a framework
• generate_dockerfile — generate optimized, production-ready Dockerfiles
• diagnose_runtime — AI-analyze runtime logs for errors and performance issues
• review_code — security, performance, and best practices code review
• suggest_rollback — analyze deployment history and recommend rollback
• summarize_alerts — AI summary of recent monitoring alerts with root cause analysis

Memory (4 tools)

• save_memory — store persistent facts for future conversations
• search_memory — search memory by keyword (with optional vector retrieval)
• list_memories — list stored memories with pagination
• delete_memory — remove a memory entry

Firewall (3 tools)

• firewall_status — get firewalld status, zones, and active rules
• list_firewall_rules — list port, service, and rich rules for a zone
• manage_firewall_rule — add or remove firewall rules

PHP (2 tools)

• php_list_runtimes — list installed PHP runtimes (FPM + FrankenPHP)
• php_list_sites — list PHP websites with version and runtime info

Caddy Management (3 tools)

• caddy_status — Caddy version, running state, config path
• caddy_restart — restart Caddy
• caddy_reload — reload configuration without downtime

Notification Channels (2 tools)

• list_notify_channels — list configured channels (Webhook, Email, Discord, Telegram)
• test_notify_channel — send a test message

Monitoring Alert Rules (3 tools)

• list_alert_rules — list all alert rules and their status
• create_alert_rule — create a threshold-based alert (e.g., CPU > 90% for 5 min)
• delete_alert_rule — remove an alert rule

App Store (3 tools)

• appstore_search_apps — search available applications
• appstore_install_app — one-click app installation via Docker
• appstore_list_installed — list installed applications

Backup (1 tool)

• trigger_backup — trigger an immediate system backup

System Inspection (1 tool)

• run_inspection — comprehensive health check: disk, memory, containers, SSL certificates, and AI-powered summary report

Cron Job Management (6 tools)

• list_cron_jobs — list all scheduled jobs (optional tag filter)
• get_cron_job_logs — get execution logs for a job
• create_cron_job — create a scheduled job (standard 5-field cron format)
• update_cron_job — update job settings
• delete_cron_job — delete a job and its logs
• trigger_cron_job — manually run a job immediately

Key Capabilities

System Health Inspection

The AI can run a comprehensive system audit checking disk usage, memory pressure, container health, SSL certificate expiration, and configuration issues. Results include an AI-generated summary with prioritized recommendations.

Self-Healing Engine

When monitoring alerts fire, the AI’s SelfHealEngine automatically analyzes the situation and suggests remediation actions. This is rate-limited (5-minute cooldown) and integrates with the monitoring alert system via EventBus.

Auto-Diagnosis on Build Failure

When a project build fails (deploy.build.failed event), the AI automatically analyzes the build logs and provides a diagnosis with suggested fixes — no manual intervention needed.

Compose Generation

Describe the application you want to deploy in natural language, and the AI generates a complete docker-compose.yml:

“Set up a WordPress site with MySQL and Redis caching”

Confirmation Dialog

Operations that modify your server trigger a confirmation dialog showing exactly what will be executed. You must approve before the action proceeds.