Docker

import { Aside } from ‘@astrojs/starlight/components’;

Quick start with Docker

git clone https://github.com/pepetox/osint-ai-one.git
cd osint-ai-one
cp .env.example .env
# Edit .env with your API keys

Available profiles

Docker Compose uses profiles to start different services:

Profile	Service	Command
cli	Interactive agent	docker compose run --rm --profile cli osint-agent
server	MCP Server (port 8080)	docker compose --profile server up osint-mcp
server	A2A Server (port 9000)	docker compose --profile server up osint-a2a
—	Ollama (port 11434)	Automatically started as dependency

Interactive CLI

docker compose run --rm --profile cli osint-agent

Opens the interactive prompt like the local installation:

osint> Investigate IP 185.220.101.34
osint> /dashboard
osint> /report save

Single query

docker compose run --rm --profile cli osint-agent -q "Investigate IP 8.8.8.8"
docker compose run --rm --profile cli osint-agent -q "Analyze domain evil.com" -- --json

MCP and A2A servers

# Start both servers in background
docker compose --profile server up -d

# View logs
docker compose logs -f osint-mcp
docker compose logs -f osint-a2a

# Stop
docker compose --profile server down

MCP will be available at http://localhost:8080/mcp and A2A at http://localhost:9000.

Volumes

Data persists between executions thanks to volumes:

Host path	Container path	Purpose
./data/	/app/data/	SQLite databases + ChromaDB
./investigaciones/	/app/investigaciones/	Investigation folder structure
./reports/	/app/reports/	Generated threat intelligence reports
./logs/	/app/logs/	Application logs
./.env	/app/.env	API keys and configuration

The Docker image pre-downloads the embeddings model (~100 MB) at build time, so investigations work offline. The Ollama LLM model is downloaded separately.

Pre-installed embeddings model

The Dockerfile includes:

RUN python -c "from sentence_transformers import SentenceTransformer; \
    SentenceTransformer('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2')"

This avoids download on first use and allows operation without HuggingFace Hub access.

Ollama in Docker

By default, the Docker image tries to connect to Ollama at http://host.docker.internal:11434 (Ollama running on your host machine).

If you want Ollama inside the container:

docker-compose.yml

services:
  ollama:
    image: ollama/ollama
    volumes:
      - ollama_data:/root/.ollama
    ports:
      - "11434:11434"

# After starting
docker compose exec ollama ollama pull qwen3:14b

Environment variables in Docker

Same as local installation, via the .env file. Docker Compose loads it automatically.

# Verify .env is loaded
docker compose run --rm --profile cli osint-agent -- osint-agent --help

Manual image build

docker build -t osint-ai-one .
docker run -it --env-file .env osint-ai-one