Skip to main content

​
Overview

SIAA (Sistema Inteligente de Apoyo Administrativo) is an AI-powered judicial document management system that uses Ollama with the Qwen2.5:3b model for intelligent document search and question answering.

​
System Requirements

Python

Python 3.8 or higher

LibreOffice

LibreOffice headless for document conversion

Tesseract

Tesseract OCR for scanned PDF processing

Ollama

Ollama with Qwen2.5:3b model

​
Installation Steps

1

Install System Dependencies

Install LibreOffice headless and Tesseract OCR on Fedora/RHEL:
sudo dnf install libreoffice-headless tesseract tesseract-langpack-spa poppler-utils -y
For Debian/Ubuntu:
sudo apt install libreoffice-headless tesseract-ocr tesseract-ocr-spa poppler-utils -y
LibreOffice headless is required for converting .doc and .pdf files to .docx format.
2

Install Python Dependencies

Install required Python packages:
pip install Flask flask-cors pandas python-docx pymupdf4llm \
            pdf2image pytesseract openpyxl xlrd requests \
            --break-system-packages
The --break-system-packages flag is needed on Fedora 43 and similar systems with externally-managed Python environments.

​
Core Dependencies

  • Flask & flask-cors: Web server and API framework
  • pandas: Excel file processing
  • python-docx: Word document parsing
  • pymupdf4llm: PDF to Markdown conversion
  • pdf2image & pytesseract: OCR for scanned documents
  • openpyxl & xlrd: Excel format support
3

Install and Configure Ollama

Install Ollama from ollama.ai:
curl -fsSL https://ollama.ai/install.sh | sh
Pull the Qwen2.5:3b model:
ollama pull qwen2.5:3b
SIAA is configured to use the qwen2.5:3b model running on http://localhost:11434. This is defined in siaa_proxy.py:182.
Verify the installation:
ollama list
You should see qwen2.5:3b in the output.
4

Create Directory Structure

Create the required SIAA directories:
sudo mkdir -p /opt/siaa/fuentes
sudo mkdir -p /opt/siaa/fuentes/normativa
sudo mkdir -p /opt/siaa/logs
sudo mkdir -p /opt/siaa/instructivos
sudo mkdir -p /opt/siaa/pdfs_origen
Set appropriate permissions:
sudo chown -R $USER:$USER /opt/siaa

​
Directory Purpose

DirectoryPurpose
/opt/siaa/fuentesConverted Markdown documents for indexing
/opt/siaa/fuentes/normativaLegal documents and norms
/opt/siaa/logsSystem logs including quality monitoring
/opt/siaa/instructivosSource Word/Excel files
/opt/siaa/pdfs_origenPDF files for conversion
5

Configure Environment Variables

Set server IP and port (optional):
export SIAA_SERVER_IP="192.168.1.100"
export SIAA_SERVER_PORT="5000"
These environment variables are read in siaa_proxy.py:168-170. If not set, the system uses the IP from the HTTP Host header and defaults to port 5000.
To make these permanent, add them to ~/.bashrc or ~/.profile:
echo 'export SIAA_SERVER_IP="192.168.1.100"' >> ~/.bashrc
echo 'export SIAA_SERVER_PORT="5000"' >> ~/.bashrc
source ~/.bashrc
6

Deploy Source Files

Copy the SIAA Python files to your deployment location:
mkdir -p ~/siaa
cp siaa_proxy.py convertidor.py convertidor_pdf.py ~/siaa/
cp -r Web ~/siaa/
The Web directory contains:
  • index.html: Main web interface
  • index2121.html: Alternative interface version
  • default.conf: Nginx configuration template

​
Configuration Details

​
Proxy Server Settings

The Flask proxy server (siaa_proxy.py) has the following key configuration: 
OLLAMA_URL             = "http://localhost:11434"
MODEL                  = "qwen2.5:3b"
CARPETA_FUENTES        = "/opt/siaa/fuentes"
LOG_ARCHIVO            = "/opt/siaa/logs/calidad.jsonl"
MAX_DOCS_CONTEXTO      = 2
CHUNK_SIZE             = 800
CHUNK_OVERLAP          = 300
MAX_CHUNKS_CONTEXTO    = 3
CACHE_MAX_ENTRADAS     = 200
CACHE_TTL_SEGUNDOS     = 3600
The cache system (LRU cache) stores up to 200 frequently-asked questions with a 1-hour TTL, providing responses in ~5ms vs 44s for uncached queries.

​
Document Converter Settings

The document converter (convertidor.py) uses these default paths: 
DEFAULT_ORIGEN  = Path("/opt/siaa/instructivos")
DEFAULT_DEST_MD = Path("/opt/siaa/fuentes")
DEFAULT_DB      = Path("/opt/siaa/institucional.db")
DEFAULT_LOG     = Path("/opt/siaa/logs/conversion_errores.log")
TEMP_DIR        = Path("/tmp/siaa_temp")

​
PDF Converter Settings

The PDF converter (convertidor_pdf.py) configuration: 
CARPETA_ENTRADA = "/opt/siaa/pdfs_origen"
CARPETA_SALIDA  = "/opt/siaa/fuentes/normativa"
MIN_CHARS = 200   # Threshold for OCR fallback
OCR_DPI   = 300
OCR_LANG  = "spa"

​
Verification

Verify your installation: 
python3 --version
# Should show Python 3.8 or higher

​
Next Steps

Quickstart Guide

Learn how to start the system and make your first query

Document Conversion

Convert your institutional documents to Markdown

​
Troubleshooting

Check if Ollama is running:
systemctl status ollama
Start Ollama if needed:
systemctl start ollama
Or run manually:
ollama serve
Ensure LibreOffice headless is properly installed:
which libreoffice
libreoffice --headless --version
Test conversion manually:
libreoffice --headless --convert-to docx --outdir /tmp test.doc
Check directory permissions:
ls -la /opt/siaa
Fix permissions:
sudo chown -R $USER:$USER /opt/siaa
chmod -R 755 /opt/siaa
Use a virtual environment to isolate dependencies:
python3 -m venv ~/siaa-env
source ~/siaa-env/bin/activate
pip install Flask flask-cors pandas python-docx pymupdf4llm pdf2image pytesseract openpyxl xlrd requests

Build docs developers (and LLMs) love

Get started for freeTalk to us