Skip to main content

Infrastructure Setup

This guide explains how to deploy the Agnox on a Linux server (Oracle Cloud / VPS).

1. Prerequisites​

Ensure the following are available before deployment:

  • Linux server with SSH access (Oracle Cloud, AWS, VPS, etc.)
  • Docker & Docker Compose installed
  • MongoDB Atlas connection string
  • Open ports for HTTP/HTTPS and RabbitMQ (if external)
  • AI API Key: At least one LLM API key is recommended for the AI Quality Orchestrator. Supported providers: Google Gemini (PLATFORM_GEMINI_API_KEY), OpenAI (PLATFORM_OPENAI_API_KEY), Anthropic (PLATFORM_ANTHROPIC_API_KEY). Organization-level BYOK keys can override platform defaults at runtime.

2. Server Configuration (.env)​

The system now supports dynamic environment mapping and AI integration.

# Database & Queue (PaaS Infrastructure)
PLATFORM_MONGO_URI=mongodb+srv://<user>:<pass>@cluster.mongodb.net/automation_platform
PLATFORM_RABBITMQ_URL=amqp://automation-rabbitmq
PLATFORM_REDIS_URL=redis://automation-redis:6379

# AI Configuration (at least one provider key recommended)
PLATFORM_GEMINI_API_KEY=<REDACTED_GOOGLE_API_KEY>
# PLATFORM_OPENAI_API_KEY=<REDACTED_OPENAI_KEY>
# PLATFORM_ANTHROPIC_API_KEY=<REDACTED_ANTHROPIC_KEY>

# Encryption (CRITICAL — exactly 32 characters / 64 hex chars)
ENCRYPTION_KEY=<64-char-hex-string>

# Auth Secrets
PLATFORM_JWT_SECRET=<REDACTED_JWT_SECRET>
PLATFORM_WORKER_CALLBACK_SECRET=<REDACTED_WORKER_SECRET>

3. The Worker Service Workflow​

The Worker is the heart of the execution engine. Here is the updated lifecycle of a test run:

  1. Pull: Worker pulls the requested Docker image (if not cached).

  2. Run: Executes the container with the secure entrypoint.sh.

  3. Stream: Logs are streamed in real-time to Redis/MongoDB and the Dashboard.

  4. Completion Check:

    • ✅ PASS: Status updated to PASSED.
    • ❌ FAIL: Status updated to ANALYZING.

  5. AI Analysis (If Failed):

    • The Worker gathers the failure logs.
    • It sends them to the configured LLM provider (Gemini, GPT-4o, or Claude — resolved via BYOK or platform fallback).
    • Generates a "Root Cause Analysis" and "Suggested Fix" using a dual-agent Analyzer → Critic pipeline.
    • Updates the execution record with the analysis report and records the aiModel used as an immutable artifact.
    • Final status set to FAILED (or UNSTABLE).

4. Deployment​

Build and start the production stack:

docker compose -f docker-compose.prod.yml up -d --build

Once deployed:

  • The Dashboard becomes accessible via browser.
  • The Producer API handles job requests.
  • The Worker listens for tasks on RabbitMQ.

5. Troubleshooting​

IssueResolution
Tests fail instantlyVerify execution config in Settings → Run Settings.
No AI AnalysisCheck if at least one AI provider key is set (PLATFORM_GEMINI_API_KEY, PLATFORM_OPENAI_API_KEY, or a BYOK key in Settings → AI Models).
Container exits immediatelyCheck entrypoint.sh permissions.
Status stuck on ANALYZINGCheck Worker logs for LLM provider timeouts or invalid API keys.
Producer service refuses to startVerify ENCRYPTION_KEY (exactly 32 chars / 64 hex), PLATFORM_JWT_SECRET, and PLATFORM_MONGO_URI are all set.