Docker Deployment Guide

This guide covers deploying Soliplex Ingester using Docker Compose for production and development environments. While it can function on its own, it is mostly intended as a guide for creating customized deplolyments to match expected configurations.

Quick Start

Starting Services

Navigate to docker directory:
```
cd docker
```
Start all services:
```
docker-compose up -d
```
Verify services are running:
```
docker-compose ps
```
Access the application:
Web UI: http://localhost:8002
API Documentation: http://localhost:8002/docs
PostgreSQL: localhost:5432
SeaweedFS: http://localhost:8333 (S3) / http://localhost:9333 (Admin)

View logs:

# All services
docker-compose logs -f

# Specific service
docker-compose logs -f soliplex_ingester

Stop services:
```
docker-compose down
```

Service Overview

Architecture Diagram

graph TB
    subgraph "Docker Network (soliplex_net)"
        Ingester["Soliplex Ingester<br/>:8000→8002<br/>(API + Worker)"]

        Postgres[("PostgreSQL<br/>:5432<br/>(Database)")]

        HAProxy["HAProxy<br/>:5004<br/>(Load Balancer)"]

        Ollama["Ollama<br/>(GPU)<br/>(Embeddings)"]

        Docling1["Docling 1<br/>(GPU)"]
        Docling2["Docling 2<br/>(GPU)"]
        Docling3["Docling 3<br/>(GPU)"]

        SeaweedFS[("SeaweedFS<br/>:8333 (S3)<br/>:9333 (Admin)")]

        Ingester -->|Stores metadata| Postgres
        Ingester -->|Parse requests| HAProxy
        Ingester -->|Embeddings| Ollama
        Ingester -.->|Optional S3| SeaweedFS

        HAProxy -->|Round-robin<br/>Cookie-based| Docling1
        HAProxy -->|Load balanced| Docling2
        HAProxy -->|Load balanced| Docling3
    end

    User([User]) -->|http://localhost:8002| Ingester

    style Ingester fill:#4a90e2,stroke:#2e5c8a,color:#fff
    style Postgres fill:#336791,stroke:#224466,color:#fff
    style HAProxy fill:#3ba13b,stroke:#2d7a2d,color:#fff
    style Ollama fill:#ff6b6b,stroke:#cc5555,color:#fff
    style Docling1 fill:#f39c12,stroke:#c27d0e,color:#fff
    style Docling2 fill:#f39c12,stroke:#c27d0e,color:#fff
    style Docling3 fill:#f39c12,stroke:#c27d0e,color:#fff
    style SeaweedFS fill:#9b59b6,stroke:#7a4691,color:#fff

Required Services

Service	Purpose	Required
soliplex_ingester	Main application (API + Worker)	Yes
postgres	Document and workflow database	Yes

Optional Services

Service	Purpose	Required
haproxy	Load balancer for Docling instances	No (but recommended for production)
docling (1-3)	PDF parsing with GPU acceleration	No (can parse without GPU or use external service)
ollama_img	Embedding generation with GPU	No (can use external embedding service)
seaweedfs	S3-compatible object storage	No (can use filesystem or cloud S3)

Resource Requirements

Minimum (Development - No GPU): - CPU: 4 cores - RAM: 8 GB - Disk: 20 GB

Recommended (Production with GPU): - CPU: 16+ cores - RAM: 64+ GB - GPU: NVIDIA GPU with 24+ GB VRAM (for Docling + Ollama) - Disk: 100+ GB

Prerequisites

Required Software

Docker Engine 20.10+
```
docker --version
```
Docker Compose 2.0+
```
docker-compose --version
```

For GPU Support

NVIDIA Container Toolkit

Install NVIDIA Container Toolkit for GPU support:

Ubuntu/Debian:

distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
curl -s -L https://nvidia.github.io/libnvidia-container/$distribution/libnvidia-container.list | \
  sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

Verify GPU access:

docker run --rm --gpus all nvidia/cuda:12.8.0-base-ubuntu22.04 nvidia-smi

Configuration

Environment Variables

Create a .env file in the docker/ directory:

# Database Configuration
POSTGRES_USER=postgres
POSTGRES_PASSWORD=your_secure_password_here
POSTGRES_DB=soliplex_attrib
DOC_DB_URL=postgresql+psycopg://soliplex_attrib:soliplex_attrib@postgres:5432/soliplex_attrib

# Storage Configuration
FILE_STORE_TARGET=fs                    # Options: fs, db, s3
FILE_STORE_DIR=/var/soliplex/file_store
LANCEDB_DIR=/var/soliplex/lancedb

# Worker Configuration
WORKER_TASK_COUNT=10                    # Number of concurrent workflow steps
INGEST_WORKER_CONCURRENCY=20            # Document ingestion concurrency

# Docling Configuration
DOCLING_SERVER_URL=http://haproxy:5004/v1
DOCLING_HTTP_TIMEOUT=1200               # Timeout in seconds
DOCLING_CONCURRENCY=4                   # Concurrent Docling requests

# Ollama Configuration (if using Ollama for embeddings)
OLLAMA_BASE_URL=http://ollama_img:11434

# SeaweedFS Configuration (if using S3 storage)
S3_ENDPOINT_URL=http://seaweedfs:8333
S3_ACCESS_KEY_ID=your_access_key
S3_SECRET_ACCESS_KEY=your_secret_key
S3_BUCKET_NAME=soliplex-artifacts

# Logging
LOG_LEVEL=INFO                          # Options: DEBUG, INFO, WARNING, ERROR

# API Configuration
API_KEY_ENABLED=false                   # Enable API key authentication

See .env.docker.example for a complete template.

Volume Management

The docker-compose configuration creates persistent volumes:

volumes:
  postgres_data:        # PostgreSQL database files
  seaweedfs_data:       # SeaweedFS object storage
  docling_artifacts:    # Docling temporary artifacts
  ollama_img_data:      # Ollama model files

Local bind mounts:

./file_store:/var/soliplex/file_store   # Document artifacts
./lancedb:/var/soliplex/lancedb         # Vector database

Backup volumes:

# Backup PostgreSQL
docker-compose exec postgres pg_dump -U postgres soliplex_attrib > backup.sql

# Backup volumes
docker run --rm -v postgres_data:/data -v $(pwd):/backup alpine tar czf /backup/postgres_data.tar.gz -C /data .

Restore volumes:

docker run --rm -v postgres_data:/data -v $(pwd):/backup alpine sh -c "cd /data && tar xzf /backup/postgres_data.tar.gz"

GPU Configuration

The docker-compose.yml configures GPU access for Docling and Ollama services.

Important: Adjust device_ids based on your hardware:

deploy:
  resources:
    reservations:
      devices:
        - driver: nvidia
          device_ids: ['3']  # ← Change this to match your GPU ID
          capabilities: [gpu]

Check available GPUs:

nvidia-smi -L

Example output:

GPU 0: NVIDIA A100-SXM4-80GB
GPU 1: NVIDIA A100-SXM4-80GB
GPU 2: NVIDIA A100-SXM4-80GB
GPU 3: NVIDIA A100-SXM4-80GB

Multiple services sharing one GPU:

The default configuration runs 3 Docling instances + Ollama all on GPU 3. This works if: - GPU has sufficient VRAM (24+ GB recommended) - Memory limits are properly configured - Workload is I/O bound (services wait on data)

Distribute across GPUs:

# Docling on GPU 0
device_ids: ['0']

# Docling_2 on GPU 1
device_ids: ['1']

# Docling_3 on GPU 2
device_ids: ['2']

# Ollama on GPU 3
device_ids: ['3']

Network Configuration

All services run on the soliplex_net bridge network for internal communication.

Port Mappings:

Host Port	Container Port	Service	Purpose
8002	8000	soliplex_ingester	Web UI & API
5432	5432	postgres	Database
5004	5004	haproxy	Docling load balancer
5000	5001	docling	Direct Docling access
5001	5001	docling_2	Direct Docling access
8333	8333	seaweedfs	S3 API
9333	9333	seaweedfs	Admin UI

Change ports if conflicts exist:

ports:
  - "8080:8000"  # Map to different host port

Service Details

Soliplex Ingester

Configuration:

soliplex_ingester:
  image: soliplex_ingester:latest
  environment:
    DOC_DB_URL: postgresql+psycopg://soliplex_attrib:soliplex_attrib@postgres:5432/soliplex_attrib
    FILE_STORE_TARGET: fs
    FILE_STORE_DIR: /var/soliplex/file_store
    LANCEDB_DIR: /var/soliplex/lancedb
    WORKER_TASK_COUNT: 10
    DOCLING_SERVER_URL: http://haproxy:5004/v1
    DOCLING_HTTP_TIMEOUT: 1200
    DOCLING_CONCURRENCY: 4
  ports:
    - "8002:8000"
  volumes:
    - ./file_store:/var/soliplex/file_store
    - ./lancedb:/var/soliplex/lancedb

Building the image:

# From project root
docker build -t soliplex_ingester:latest .

Scaling workers:

The default configuration runs the API server with an integrated worker. For production, separate them:

# API Server
soliplex_api:
  image: soliplex_ingester:latest
  command: si-cli serve --host 0.0.0.0 --workers 4
  # ... rest of config

# Workers (scale as needed)
soliplex_worker:
  image: soliplex_ingester:latest
  command: si-cli worker
  deploy:
    replicas: 3
  # ... rest of config

PostgreSQL

Configuration:

postgres:
  image: postgres:18-trixie
  environment:
    POSTGRES_USER: postgres
    POSTGRES_PASSWORD: postgres
    POSTGRES_INITDB_ARGS: "-A scram-sha-256"
  ports:
    - "5432:5432"
  volumes:
    - postgres_data:/var/lib/postgresql
    - ./pgsql/config/init.sql:/docker-entrypoint-initdb.d/init.sql

Initialization Script:

The init.sql script creates users and databases:

-- Create application user
CREATE USER soliplex_attrib WITH PASSWORD 'soliplex_attrib';

-- Create database
CREATE DATABASE soliplex_attrib OWNER soliplex_attrib;

-- Grant permissions
GRANT ALL PRIVILEGES ON DATABASE soliplex_attrib TO soliplex_attrib;

Production Security:

⚠️ The example uses weak passwords for development. For production:

Use strong, randomly generated passwords
Store credentials in Docker secrets or environment files
Restrict network access
Enable SSL/TLS connections

Connection string format:

postgresql+psycopg://user:password@host:port/database

Docling Services

Docling converts PDF documents to markdown and structured JSON.

Configuration (per instance):

docling:
  image: ghcr.io/docling-project/docling-serve-cu128
  environment:
    DOCLING_SERVE_ENG_LOC_NUM_WORKERS: 4
    DOCLING_SERVE_ARTIFACTS_PATH: "/artifacts"
    DOCLING_NUM_THREADS: 16
    UVICORN_WORKERS: 1
    PYTORCH_CUDA_ALLOC_CONF: "expandable_segments:True"
    DOCLING_SERVE_ENABLE_UI: 1
    DOCLING_SERVE_MAX_SYNC_WAIT: 9999
    NVIDIA_VISIBLE_DEVICES: "all"
    DOCLING_SERVE_ENABLE_REMOTE_SERVICES: True
  restart: "unless-stopped"
  runtime: "nvidia"
  volumes:
    - docling_artifacts:/artifacts
  deploy:
    resources:
      limits:
        memory: 32000M
      reservations:
        devices:
          - driver: nvidia
            device_ids: ['3']
            capabilities: [gpu]

Key Environment Variables:

DOCLING_SERVE_ENG_LOC_NUM_WORKERS: 4 - Layout analysis workers
DOCLING_NUM_THREADS: 16 - Processing threads
PYTORCH_CUDA_ALLOC_CONF: "expandable_segments:True" - GPU memory optimization
DOCLING_SERVE_MAX_SYNC_WAIT: 9999 - Long timeout for large documents

Memory Management:

⚠️ Docling is prone to memory leaks with long-running processes.

Mitigation strategies:

Memory limits: memory: 32000M prevents OOM killing other services
Restart policy: restart: unless-stopped recovers from crashes
Load balancing: Multiple instances with HAProxy provide redundancy
Health checks: HAProxy routes around unhealthy instances

Without GPU:

Remove GPU configuration and use CPU-only image:

docling:
  image: ghcr.io/docling-project/docling-serve  # CPU-only
  runtime: null  # Remove nvidia runtime
  deploy:
    resources:
      limits:
        memory: 16000M
      # Remove GPU reservation

HAProxy Load Balancer

HAProxy distributes requests across multiple Docling instances.

Configuration:

haproxy:
  image: docker.io/library/haproxy:3.3-alpine
  ports:
    - 5004:5004
  volumes:
    - ./haproxy/haproxy.cfg:/usr/local/etc/haproxy/haproxy.cfg

HAProxy Configuration (haproxy/haproxy.cfg):

global
    maxconn 256

defaults
    mode http
    timeout connect 5000ms
    timeout client 1200000ms
    timeout server 1200000ms

frontend docling_frontend
    bind *:5004
    default_backend docling_backend

backend docling_backend
    balance roundrobin
    cookie SERVERID insert indirect nocache
    server docling1 docling:5001 check cookie docling1
    server docling2 docling_2:5001 check cookie docling2
    server docling3 docling_3:5001 check cookie docling3

Load Balancing Strategy:

Round-robin: Distributes requests evenly
Cookie-based persistence: Same client goes to same server
Health checks: Removes failed servers from rotation

Why cookie-based persistence?

Docling parsing is stateful for multi-step conversions. The ingester client uses cookies to ensure all requests for a document go to the same Docling instance.

Ollama

Ollama provides embedding generation for vector search.

Configuration:

ollama_img:
  image: ollama/ollama:latest
  container_name: ollama_img
  volumes:
    - ollama_img_data:/root/.ollama
  restart: always
  deploy:
    resources:
      limits:
        memory: 32000M
      reservations:
        devices:
          - driver: nvidia
            device_ids: ['3']
            capabilities: [gpu]

Pull models:

docker-compose exec ollama_img ollama pull nomic-embed-text
docker-compose exec ollama_img ollama pull llama2

List installed models:

docker-compose exec ollama_img ollama list

Ingester Configuration:

Set in .env:

OLLAMA_BASE_URL=http://ollama_img:11434

SeaweedFS

SeaweedFS provides S3-compatible object storage for artifacts.

Configuration:

seaweedfs:
  image: ghcr.io/chrislusf/seaweedfs
  command: server -s3 -s3.config=/config/config.json
  ports:
    - 8333:8333  # S3 API
    - 9333:9333  # Admin UI
  volumes:
    - seaweedfs_data:/data
    - ./seaweedfs/config:/config

Initialization:

The seaweedfs-init container runs setup:

seaweedfs-init:
  image: ghcr.io/chrislusf/seaweedfs
  entrypoint: ["/bin/sh"]
  volumes:
    - ./seaweedfs/config/init.sh:/init.sh
  command: ["/init.sh"]

Example init.sh:

#!/bin/sh
# Wait for SeaweedFS to start
sleep 5

# Create bucket
curl -X PUT http://seaweedfs:8333/soliplex-artifacts

# Configure credentials (in config.json)

Ingester Configuration:

Set in .env:

FILE_STORE_TARGET=s3
S3_ENDPOINT_URL=http://seaweedfs:8333
S3_ACCESS_KEY_ID=your_access_key
S3_SECRET_ACCESS_KEY=your_secret_key
S3_BUCKET_NAME=soliplex-artifacts

Alternative: Use Cloud S3

FILE_STORE_TARGET=s3
S3_ENDPOINT_URL=https://s3.amazonaws.com
S3_ACCESS_KEY_ID=your_aws_key
S3_SECRET_ACCESS_KEY=your_aws_secret
S3_BUCKET_NAME=your-bucket-name
AWS_REGION=us-east-1

Authentication Setup

For production deployments with authentication, use docker-compose.auth.yml.

OAuth2 Proxy Stack

The auth configuration adds: - NGINX - Reverse proxy with SSL termination - OAuth2 Proxy - OIDC authentication - Soliplex Ingester - Configured to trust proxy headers

Start with authentication:

docker-compose -f docker-compose.yml -f docker-compose.auth.yml up -d

Configuration

Create .env.auth file:

See docker/.env.auth.example:

# OIDC Provider Configuration
OAUTH2_PROVIDER=oidc
OAUTH2_OIDC_ISSUER_URL=https://your-oidc-provider.com
OAUTH2_CLIENT_ID=your_client_id
OAUTH2_CLIENT_SECRET=your_client_secret
OAUTH2_REDIRECT_URL=https://your-domain.com/oauth2/callback

# OAuth2 Proxy Configuration
OAUTH2_COOKIE_SECRET=random_32_char_secret_here
OAUTH2_COOKIE_DOMAIN=your-domain.com

# Soliplex Ingester Configuration
AUTH_TRUST_PROXY_HEADERS=true
AUTH_USER_HEADER=X-Forwarded-User
AUTH_EMAIL_HEADER=X-Forwarded-Email

Configure NGINX:

Edit docker/nginx/nginx.conf for your domain and SSL certificates.

Configure OAuth2 Proxy:

Edit docker/oauth2-proxy/oauth2-proxy.cfg for your OIDC provider.

See AUTHENTICATION.md for detailed setup instructions.

Production Deployment

Security Best Practices

Use Strong Passwords:

# Generate secure password
openssl rand -base64 32

Use Docker Secrets:

secrets:
  db_password:
    file: ./secrets/db_password.txt

services:
  postgres:
    secrets:
      - db_password
    environment:
      POSTGRES_PASSWORD_FILE: /run/secrets/db_password

Restrict Network Access:

services:
  postgres:
    ports: []  # Don't expose to host
    networks:
      - soliplex_net  # Internal only

Enable SSL/TLS:

Use NGINX with Let's Encrypt certificates or cloud load balancer.

Scan Images:
```
docker scan soliplex_ingester:latest
```

Scaling Configuration

For high throughput:

services:
  soliplex_api:
    command: si-cli serve --host 0.0.0.0 --workers 8
    deploy:
      replicas: 2

  soliplex_worker:
    command: si-cli worker
    environment:
      WORKER_TASK_COUNT: 20
    deploy:
      replicas: 5

  docling:
    deploy:
      replicas: 5
      resources:
        limits:
          memory: 24000M

Resource Limits

Set appropriate limits to prevent resource exhaustion:

services:
  soliplex_ingester:
    deploy:
      resources:
        limits:
          cpus: '4.0'
          memory: 8G
        reservations:
          cpus: '2.0'
          memory: 4G

Health Checks

Add health checks for automatic recovery:

services:
  soliplex_ingester:
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/api/v1/batch/"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

Logging

Configure logging drivers:

services:
  soliplex_ingester:
    logging:
      driver: "json-file"
      options:
        max-size: "100m"
        max-file: "5"

Or use external logging:

logging:
  driver: "syslog"
  options:
    syslog-address: "tcp://192.168.0.1:514"

Monitoring and Maintenance

Health Checks

Check service status:

docker-compose ps

View resource usage:

docker stats

Check specific service:

docker-compose exec soliplex_ingester curl http://localhost:8000/api/v1/stats/durations

Log Management

View logs:

# All services
docker-compose logs -f

# Specific service
docker-compose logs -f soliplex_ingester

# Last 100 lines
docker-compose logs --tail=100 docling

# Since timestamp
docker-compose logs --since 2026-01-22T10:00:00

Export logs:

docker-compose logs > logs.txt

Database Maintenance

Backup:

docker-compose exec postgres pg_dump -U postgres soliplex_attrib | gzip > backup_$(date +%Y%m%d).sql.gz

Restore:

gunzip < backup_20260122.sql.gz | docker-compose exec -T postgres psql -U postgres soliplex_attrib

Vacuum database:

docker-compose exec postgres psql -U postgres -d soliplex_attrib -c "VACUUM ANALYZE;"

Vector Database Maintenance

Vacuum LanceDB:

curl "http://localhost:8002/api/v1/lancedb/vacuum?db=default"

Check database size:

docker-compose exec soliplex_ingester du -sh /var/soliplex/lancedb

Updates

Update images:

docker-compose pull
docker-compose up -d

Update single service:

docker-compose pull soliplex_ingester
docker-compose up -d soliplex_ingester

Rebuild custom image:

docker build -t soliplex_ingester:latest .
docker-compose up -d soliplex_ingester

Troubleshooting

Common Issues

Services Won't Start

Check logs:

docker-compose logs <service_name>

Check configuration:

docker-compose config

Validate environment variables:

docker-compose exec soliplex_ingester env | grep DOC_DB_URL

Database Connection Errors

Error: connection to server at "postgres" (172.18.0.2), port 5432 failed

Solutions: 1. Verify postgres is running: docker-compose ps postgres 2. Check network: docker-compose exec soliplex_ingester ping postgres 3. Verify credentials in DOC_DB_URL 4. Check postgres logs: docker-compose logs postgres

Test connection:

docker-compose exec soliplex_ingester psql "postgresql://soliplex_attrib:soliplex_attrib@postgres:5432/soliplex_attrib"

GPU Not Available

Error: could not select device driver "" with capabilities: [[gpu]]

Solutions: 1. Install NVIDIA Container Toolkit (see Prerequisites) 2. Restart Docker daemon: sudo systemctl restart docker 3. Verify GPU access: docker run --rm --gpus all nvidia/cuda:12.8.0-base-ubuntu22.04 nvidia-smi

Check GPU usage:

docker-compose exec docling nvidia-smi

Docling Memory Errors

Error: CUDA out of memory or container crashes

Solutions: 1. Reduce DOCLING_NUM_THREADS in docker-compose.yml 2. Reduce DOCLING_SERVE_ENG_LOC_NUM_WORKERS 3. Increase GPU memory by using dedicated GPU per instance 4. Process smaller documents or reduce concurrency

Monitor memory:

docker stats docling

Port Conflicts

Error: bind: address already in use

Solutions: 1. Check what's using the port:

# Linux/macOS
lsof -i :8002

# Windows
netstat -ano | findstr :8002

Change port mapping in docker-compose.yml:

ports:
  - "8080:8000"  # Use different host port

Volume Permission Errors

Error: permission denied when accessing volumes

Solutions: 1. Check ownership:

ls -la ./file_store ./lancedb

Fix permissions:

sudo chown -R $(id -u):$(id -g) ./file_store ./lancedb

Or use Docker's user mapping:
```
user: "${UID}:${GID}"
```

SeaweedFS Connection Errors

Error: Unable to connect to S3 endpoint

Solutions: 1. Verify SeaweedFS is running: docker-compose ps seaweedfs 2. Check initialization completed: docker-compose logs seaweedfs-init 3. Test S3 endpoint:

curl http://localhost:8333

Verify bucket exists:

docker-compose exec seaweedfs weed shell
> s3.bucket.list

HAProxy Routing Issues

Symptom: Requests timing out or routing to unhealthy Docling instances

Debug: 1. Check HAProxy stats (if enabled):

http://localhost:5004/stats

View HAProxy logs:
```
docker-compose logs haproxy
```

Test Docling instances directly:

curl http://localhost:5000/v1/health
curl http://localhost:5001/v1/health

Restart HAProxy:
```
docker-compose restart haproxy
```

Performance Issues

Slow Document Processing

Diagnose:

# Check worker activity
curl http://localhost:8002/api/v1/workflow/steps?status=RUNNING

# Check CPU/memory
docker stats

# Check Docling queue
docker-compose logs docling | grep -i queue

Solutions: 1. Increase worker concurrency: WORKER_TASK_COUNT=20 2. Add more Docling instances 3. Increase DOCLING_CONCURRENCY 4. Scale workers: docker-compose up -d --scale soliplex_worker=5

High Memory Usage

Monitor:

docker stats --no-stream

Solutions: 1. Reduce concurrency settings 2. Increase memory limits 3. Add swap space (not recommended for production) 4. Use smaller batch sizes

Disk Space Issues

Check usage:

docker system df

Clean up:

# Remove unused images
docker image prune -a

# Remove stopped containers
docker container prune

# Remove unused volumes
docker volume prune

# Clean everything
docker system prune -a --volumes

Recovery Procedures

Reset Everything

⚠️ This deletes all data!

docker-compose down -v
rm -rf ./file_store/* ./lancedb/*
docker-compose up -d

Reset Database Only

docker-compose down
docker volume rm docker_postgres_data
docker-compose up -d postgres
# Wait for postgres to initialize
docker-compose exec postgres psql -U postgres -f /docker-entrypoint-initdb.d/init.sql
docker-compose up -d

Restart Single Service

docker-compose restart soliplex_ingester

Force Recreate Service

docker-compose up -d --force-recreate soliplex_ingester

Additional Resources

Soliplex Ingester Documentation: ../README.md
API Reference: API.md
Configuration Guide: CONFIGURATION.md
Authentication Guide: AUTHENTICATION.md
Docling Documentation: https://docling-project.github.io/docling/
Docker Compose Documentation: https://docs.docker.com/compose/
NVIDIA Container Toolkit: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/

Getting Help

If you encounter issues not covered in this guide:

Check the main troubleshooting guide
Review service-specific logs
Open an issue on GitHub with:
Output of docker-compose ps
Relevant logs from docker-compose logs
Your docker-compose.yml modifications
Environment details (OS, Docker version, GPU info)

Docker Deployment Guide

Table of Contents

Quick Start

Starting Services

Service Overview

Architecture Diagram

Required Services

Optional Services

Resource Requirements

Prerequisites

Required Software

For GPU Support

Configuration

Environment Variables

Volume Management

GPU Configuration

Network Configuration

Service Details

Soliplex Ingester

PostgreSQL

Docling Services

HAProxy Load Balancer

Ollama

SeaweedFS

Authentication Setup

OAuth2 Proxy Stack

Configuration

Production Deployment

Security Best Practices

Scaling Configuration

Resource Limits

Health Checks

Logging

Monitoring and Maintenance

Health Checks

Log Management

Database Maintenance

Vector Database Maintenance

Updates

Troubleshooting

Common Issues

Services Won't Start

Database Connection Errors

GPU Not Available

Docling Memory Errors

Port Conflicts

Volume Permission Errors

SeaweedFS Connection Errors

HAProxy Routing Issues

Performance Issues

Slow Document Processing

High Memory Usage

Disk Space Issues

Recovery Procedures

Reset Everything

Reset Database Only

Restart Single Service

Force Recreate Service

Additional Resources

Getting Help