netiob 0.1.38

NET IOB

NetIOB (netiob) Calculation and Blood Glucose Prediction Library

netiob is an advanced Python package for automated blood glucose prediction and insulin-on-board (IOB) calculation built upon the OpenAPS oref0 reference algorithm. This toolkit enables researchers, data scientists, and developers to preprocess diabetes management data, compute net IOB, forecast glucose, and visually analyze prediction scenarios.

Core Features

• Blood Glucose (BG) Prediction: End-to-end pipeline using the OpenAPS oref0 determine-basal algorithm.
• Net Insulin On Board (IOB): Robust IOB calculation incorporating basal rate variations, boluses, and autosensitivity analysis.
• Data Preprocessing Utilities: Transform raw pump, CGM, and nutrition logs into SDTM-compliant, validated formats.
• Prediction Graphs: Inbuilt support (via Matplotlib/Plotly) for visualizing BG forecasting results and insulin activity.
• Core Utilities: Comprehensive API bridge, datetime handling, profile assembly, autosensitivity computation, and more.

Requirements

• Python >= 3.12, < 3.13
• NumPy >= 1.22.0
• Pandas >= 1.3.0
• Requests >= 2.25.0
• python-decouple >= 3.6
• Plotly >= 5.9.0
• A running OpenAPS oref0 HTTP API server (remote endpoint for the core IOB / determine-basal calculations)

The single source of truth for dependencies is pyproject.toml; resolved versions are pinned in uv.lock.

• Runtime configuration (read by netiob/settings.py via python-decouple):

  • OREF0_API_SERVER_URL — base URL of the oref0 HTTP server (default http://localhost:3000).
  • MAX_WORKERS — thread-pool size for the parallel net-IOB calculator (default 4).

• oref0 dependency: netiob does not implement the IOB / determine-basal math itself — it preprocesses data into oref0-compatible payloads and calls a running oref0 server, which performs the core JavaScript computations. A reachable oref0 server is therefore a hard runtime dependency. Get the oref0 repo here: https://github.com/openaps/oref0/tree/master

Installation

PyPI

pip install netiob

From source (with uv)

This project is managed with uv. Clone the repo and sync the managed virtualenv to uv.lock:

git clone https://gitlab.com/CeADARIreland/UCD/con/netiob-package.git
cd netiob
uv sync                            # create/update .venv to match uv.lock
uv run python -c "import netiob"   # run anything inside the managed env

From source (with pip)

git clone https://gitlab.com/CeADARIreland/UCD/con/netiob-package.git
cd netiob
pip install .

Makefile commands

The project ships a Makefile wrapping the common uv workflows.

Command	Description
make setup-local	Sync the managed virtualenv to uv.lock (uv sync).
make run-tests-all	Run the full deterministic suite (unit + contract + cross-cutting). Integration tests are deselected so the run needs no external services.
make run-tests-integration	Run only the integration tier against a live oref0 server. Tests self-skip if OREF0_API_SERVER_URL is unreachable.
make build-dist	Bump the version (BUMP=patch by default) and build a wheel.
make build-dist-publish	Clean dist/, rebuild, and upload to PyPI via twine.

make run-tests-all                 # fast, no network — use as the default gate
make run-tests-integration         # requires a reachable oref0 server
make build-dist BUMP=minor         # override the version bump segment

Test tiers and pass/fail criteria are described in tests/TEST_PLAN.md; worked input/output vectors for the computational core are in tests/TEST_DATA.md.

Import

# Core modules
from netiob.utils.coreutils import *
from netiob.utils.calculators import calculate_net_iob
from netiob.utils.openapsprediction import OpenapsPredictor
from netiob.utils.preprocessors import preprocess_basal_data, preprocess_bolus_data, preprocess_carbs_data, preprocess_cgm_data, preprocess_insulin_data
from netiob.utils.cgmdataprocessor import CGMDataProcessor
from netiob.utils.bgpredgraph import bg_prediction_graph

Functionalities

1. Core Utilities (coreutils.py)

• API Communication: call_api(endpoint, payload)
• Datetime Handling: parse_iso_utc, to_iso_z, etc.
• Profile Construction: get_profile_inputs, get_profile_settings
• Pump History Utilities: get_pump_history, get_carb_history
• Autosensitivity Calculation: get_autosens
• Data Serialization: make_json_serializable()

2. NetIOB Calculator (calculators.py)

• Parallel net IOB calculation at 5-min intervals via OpenAPS oref0 API
• Handles both bolus and basal variabilities (including temp basals vs. scheduled)
• Example usage:

from netiob.utils.calculators import calculatenetiob
iob_series = calculate_net_iob(basal_df, bolus_df, cgm_df, profile_df, autosens)

• Returns a list of dicts for each time point, including keys: iob, activity, basaliob, bolusiob, time, lastTemp

3. Preprocessors (preprocessors.py)

• Preprocesses CGM, basal, bolus, and meal data into standardized DataFrames
• Chunks long basal records; distributes extended boluses; aligns with OpenAPS SDTM data conventions
• Entry points: preprocess_basal_data(basal_df), preprocess_bolus_data(bolus_df), preprocess_cgm_data(cgm_df), preprocess_carbs_data(carbs_df)
• Entry point to process insulin data ready for netiob calculation and prediction: preprocess_insulin_data(processed_basal, processed_bolus)
• Entry point to process and access all processed data at once: preprocess_user_data(basal_df, bolus_df, carbs_df, cgm_df). Returns a tuple of 5 (objects) processed data.

4. CGM Data Processor (cgmdataprocessor.py)

• Orchestrates the transformation of raw user data into oref0-compatible input using the preprocessor.py
• Access processed DataFrames and prediction-ready structures:
• Pools and provides netiob and prediction required data as objects

processor = CGMDataProcessor(basal_df, bolus_df, carbs_df, cgm_df, profile_df)
processed_glucose_data = processor.glucose_data  # ready for API call
processed_basal = processor.proc_basal

5. Blood Glucose Prediction (openapsprediction.py)

• OpenapsPredictor: Automated pipeline for forecasting BG, carbs-on-board (COB), IOB, insulin needs, and clinical recommendations using oref0 APIs

from netiob.utils.openapsprediction import OpenapsPredictor
predictor = OpenapsPredictor(basal_df, bolus_df, carbs_df, cgm_df, profile_df)
result = predictor.predict_bg(currenttemp, clock, simcarb)

• Returns comprehensive prediction dict: eventualBG, predBGs, rate, reason, and more

6. Graph Plotting (bgpredgraph.py)

• Plot BG prediction scenarios:

from netiob.utils.bgpredgraph import bg_prediction_graph
fig = bg_prediction_graph(result, scenario_label="BG Prediction", fig_show=True)

• Generates interactive or static charts using Plotly/Matplotlib, highlighting:
  • IOB activity
  • COB/UAM scenarios
  • Target BG lines
  • Safest prediction curves

Example Pipeline Usage

• Import and Load Dataframes

import pandas as pd
from netiob.utils.cgmdataprocessor import CGMDataProcessor
from netiob.utils.openapsprediction import OpenapsPredictor
from netiob.utils.bgpredgraph import bg_prediction_graph
from netiob.utils.calculators import calculate_net_iob

# Load user data into DataFrames
# Note: Data can be transformed from any data sources. Only ensure they are in DataFrame format
basal_df = pd.read_csv("basal.csv")
bolus_df = pd.read_csv("bolus.csv")
carbs_df = pd.read_csv("carbs.csv")
cgm_df = pd.read_csv("cgm.csv")
profile_df = pd.read_csv("profile.csv")

• Preprocess

# Create an instance of CGMDataProcessor.
# See code file for full documentation on all objects of CGMDataProcessor
dataprocessor = CGMDataProcessor(basal_df, bolus_df, carbs_df, cgm_df, profile_df=profile)

# Get the objects of dataprocessor (Below are all the objects of CGMDataProcessor)
dataprocessor.proc_basal        # Processed basal data 
dataprocessor.proc_bolus        # Processed bolus data
dataprocessor.proc_carbs        # Processed carbs data
dataprocessor.proc_glucose      # Processed glucose data
dataprocessor.proc_insulin      # processed insulin data
dataprocessor.glucose_data      # Glucosed data structured in the required dict for OpenAPS prediction
dataprocessor.clock             # CGM last data event timestamp (for prediction)
dataprocessor.pumphistory       # Insulin data structured in the required dict (for netiob calculation and prediction)
dataprocessor.pump_clock        # Last insulin data event timestamp
dataprocessor.settings          # Useer settings in structured dict for prediction
dataprocessor.bg_targets        # BG target dict extracted from settings
dataprocessor.basal_profile     # 24-hours (hourly) baseline user basal profile dict extrapolated from user profile settings
dataprocessor.sensitivities     # Sensitivities data extracted from profile settings
dataprocessor.profile_carbs     # Carbs dict based on user profile
dataprocessor.profile           # Entire user profile structure dict needed for netiob calculation and prediction
datprocessor.autosens           # autosensitivity ratio dict calculated using oref0 API

• Prediction

predictor = OpenapsPredictor(proc_basal, proc_bolus, proc_carbs, proc_cgm, profile_df)
prediction_result = predictor.predict_bg(currenttemp={}, clock="2025-11-07T16:00:00Z")
    
# Note: Synthetic carbs entry (in grams) can be simulated for prediction. In such case, pass (e.g., 5g) carbs as thus:
prediction_result = predictor.predict_bg(currenttemp={}, clock="2025-11-07T16:00:00Z", sim_carb=5)

• Visualization

bg_prediction_graph(prediction_result, scenario_label="BG Prediction After Meal")

• Calculate NetIOB

# Note, profile_df and autosens can be None. 
# In such case, a default baseline basal profile will be calculated based on insulin history and default auto sensitivity ratio will be used.
calculate_net_iob(basal_df, bolus_df, carbs_df, cgm_df, profile_df, austosens={})

Graph Output

• The bg_prediction_graph function visualizes predicted curves (IOB, COB/UAM, ZT) and overlays "eventual BG" markers and target ranges.

Contributing

• Create issues or merge requests on GitLab for improvements.
• Follow coding conventions and update docstrings.

License

This project is licensed under the Apache License 2.0. For more details, please see the LICENSE file in the repository.

---

Maintainers

CeADAR - Ireland's Centre for AI