OCI Generative AI Service (example: CSV report graphing)

$ ask-oci-for-graph.py -h

usage: ask-oci-for-graph.py [-h] [--project-id PROJECT_ID] [--api-key API_KEY] [--region REGION] [--model MODEL]
                                             csv_file

Ask OCI Gen AI to create graphs from CSV

positional arguments:
  csv_file              Path to combined CSV file

optional arguments:
  -h, --help            show this help message and exit
  --project-id PROJECT_ID
  --api-key API_KEY
  --region REGION
  --model MODEL

#!/usr/bin/env python3

import argparse
import os
import sys
from pathlib import Path

from openai import OpenAI

def parse_args():
    parser = argparse.ArgumentParser(
        description="Ask OCI Gen AI to create graphs from CSV"
    )
    parser.add_argument("csv_file", help="Path to combined CSV file")
    parser.add_argument("--project-id", default=os.environ.get("OCI_GENAI_PROJECT_ID"))
    parser.add_argument("--api-key", default=os.environ.get("OCI_GENAI_API_KEY"))
    parser.add_argument("--region", default="us-phoenix-1")
    parser.add_argument("--model", default=os.environ.get("OCI_GENAI_MODEL", "google.gemini-2.5-flash"))
    return parser.parse_args()


def make_client(region, project_id, api_key):
    return OpenAI(
        base_url=f"https://inference.generativeai.{region}.oci.oraclecloud.com/openai/v1",
        api_key=api_key,
        project=project_id,
    )


def read_csv_text(csv_path):
    with open(csv_path, "r", encoding="utf-8") as f:
        return f.read()


def save_text(path, text):
    path.write_text(text or "", encoding="utf-8")


def main():
    args = parse_args()

    csv_path = Path(args.csv_file).resolve()
    outdir = csv_path.parent

    if not csv_path.exists():
        print(f"CSV file not found: {csv_path}", file=sys.stderr)
        return 1

    if not args.project_id:
        print("Missing --project-id or OCI_GENAI_PROJECT_ID", file=sys.stderr)
        return 1

    if not args.api_key:
        print("Missing --api-key or OCI_GENAI_API_KEY", file=sys.stderr)
        return 1

    print("Reading CSV...")
    csv_text = read_csv_text(csv_path)

    print("Creating OCI client...")
    client = make_client(args.region, args.project_id, args.api_key)

    summary_prompt = f"""
Read this CSV data.

The file may contain duplicate header rows because it may have been combined with plain cat.
Ignore repeated header rows like:
Site,OS Type,User,Mgr1,Mgr2,Mgr3,Installs

Expected columns:
- Site
- OS Type
- User
- Mgr1
- Mgr2
- Mgr3
- Installs

Return only a concise summary with:
1. total installs by Site
2. total installs by OS Type
3. top 10 users by Installs
4. manager chain Mgr1, Mgr2, Mgr3 for the top 10 users

CSV data:

{csv_text}
"""

    print("Request 1: summary...")
    summary_response = client.responses.create(
        model=args.model,
        input=summary_prompt,
    )

    summary_text = getattr(summary_response, "output_text", "") or ""
    summary_file = outdir / "oci_summary.txt"
    save_text(summary_file, summary_text)
    print(f"Saved: {summary_file}")

    chart_prompt = f"""
Read this CSV data.

The file may contain duplicate header rows because it may have been combined with plain cat.
Ignore repeated header rows like:
Site,OS Type,User,Mgr1,Mgr2,Mgr3,Installs

Expected columns:
- Site
- OS Type
- User
- Mgr1
- Mgr2
- Mgr3
- Installs

Use Python to parse the CSV data and create:
1. a stacked bar chart by Site split into Linux and Solaris/AK
2. a horizontal bar chart of top 10 users (like User,Mgr1,Mgr2,Mgr3) by Installs
3. a short summary
4. Save the charts as PNG files

If file outputs are supported, save charts as PNG files and mention the filenames.
Again, we wan't User,Mgr1,Mgr2,Mgr3 on PNG files.

CSV data:

{csv_text}
"""

    print("Request 2: charts...")
    chart_response = client.responses.create(
        model=args.model,
        tools=[{"type": "code_interpreter"}],
        instructions="Use Python to analyze the CSV text and generate charts.",
        input=chart_prompt,
    )

    chart_text = getattr(chart_response, "output_text", "") or ""
    chart_file = outdir / "oci_charts.txt"
    save_text(chart_file, chart_text)
    print(f"Saved: {chart_file}")

    print("Done.")
    return 0


if __name__ == "__main__":
    sys.exit(main())

#!/usr/bin/env python3
import pandas as pd
import io
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker

# Provided CSV data
csv_data = """Host,OS,User,Mgr1,Mgr2,Mgr3,Installs
hostname-1,Linux,user-1,user-1,user-3,111
hostname-1,Linux,user-a,user-b,user-c,22
hostname-1,Ubuntu,user-aa,user-bb,user-cc,3
"""

# Read the CSV data into a pandas DataFrame
df = pd.read_csv(io.StringIO(csv_data))

# Remove rows that are duplicates of the header.
# This filters out rows where the 'Site' column is literally 'Site',
# indicating a repeated header that was read as data.
df = df[df['Site'] != 'Site'].copy()

# Convert 'Installs' column to numeric.
# Errors are coerced to NaN, and then rows with NaN in 'Installs' are dropped.
# Finally, convert to integer type.
df['Installs'] = pd.to_numeric(df['Installs'], errors='coerce')
df.dropna(subset=['Installs'], inplace=True)
df['Installs'] = df['Installs'].astype(int)

# --- 1. Stacked bar chart by Site split into Linux and Solaris/AK ---
# Group by 'Site' and 'OS Type' and sum 'Installs'
site_os_installs = df.groupby(['Site', 'OS Type'])['Installs'].sum().unstack(fill_value=0)

# Ensure both 'Linux' and 'Solaris/AK' columns exist for consistent plotting,
# even if one OS type has no data for a given site.
for os_type in ['Linux', 'Solaris/AK']:
    if os_type not in site_os_installs.columns:
        site_os_installs[os_type] = 0

# Reorder columns to ensure consistent stacking order in the plot
site_os_installs = site_os_installs[['Linux', 'Solaris/AK']]

plt.figure(figsize=(12, 7))
site_os_installs.plot(kind='bar', stacked=True, figsize=(12, 7))
plt.title('Total Installs by Site and OS Type')
plt.xlabel('Site')
plt.ylabel('Total Installs')
plt.xticks(rotation=45, ha='right')
plt.legend(title='OS Type')
plt.tight_layout()
stacked_bar_chart_filename = 'installs_by_site_os_stacked_bar_chart.png'
plt.savefig(stacked_bar_chart_filename)
plt.clf() # Clear the current figure to prevent plots from overlapping

# --- 2. Horizontal bar chart of top 10 users (User,Mgr1,Mgr2,Mgr3) by Installs ---
# Create a combined 'User Group' column as requested
df['UserGroup'] = df['User'] + ',' + df['Mgr1'] + ',' + df['Mgr2'] + ',' + df['Mgr3']

# Group by 'UserGroup' and sum 'Installs'
user_installs = df.groupby('UserGroup')['Installs'].sum()

# Get the top 10 user groups by installs
top_10_users = user_installs.nlargest(10)

plt.figure(figsize=(12, 8))
# Sort values for the horizontal bar chart to have the largest bar at the top
top_10_users.sort_values().plot(kind='barh', color='skyblue')
plt.title('Top 10 User Groups by Installs')
plt.xlabel('Total Installs')
plt.ylabel('User Group (User, Mgr1, Mgr2, Mgr3)')
plt.tight_layout()
top_10_users_bar_chart_filename = 'top_10_users_by_installs_horizontal_bar_chart.png'
plt.savefig(top_10_users_bar_chart_filename)
plt.clf() # Clear the current figure

# --- 3. Short summary ---
total_installs = df['Installs'].sum()
unique_sites = df['Site'].nunique()
unique_os_types = df['OS Type'].nunique()
unique_users = df['UserGroup'].nunique()

summary_output = f"""
Summary of Installs Data:
----------------------------
Total installs recorded: {total_installs}
Number of unique sites: {unique_sites}
Number of unique OS types: {unique_os_types}
Number of unique user groups: {unique_users}

The data provides insights into software installations, showing their distribution
across different sites and operating system types, and identifies the top contributing
user groups based on the combined 'User,Mgr1,Mgr2,Mgr3' identifier.
"""

# Output the results
print(summary_output)
print(f"Stacked bar chart saved as: {stacked_bar_chart_filename}")
print(f"Horizontal bar chart saved as: {top_10_users_bar_chart_filename}")