November 2026
Every post published in November 2026, newest first. Pick a day to jump straight to it.
- Posts
- 802
- Topics
- 75
- Reading time
- 91 h
- Active days
- 15
Posts 31–60 of 802 · November 19, 2026
Page 2 of 27
Thursday, November 19, 2026
30 of 57 posts
How to Set Microsoft.Identity.Web Log Levels in ASP.NET Core
An API protected with Microsoft.Identity.Web logs four Information lines for every valid token. Set Microsoft.Identity to Warning in appsettings.json to keep only failures, and bring the detail back in Development when you investigate a refused token. Real logs from .NET 10 and Microsoft.Identity.Web 4.18.0.
9 min readHow to Protect an ASP.NET Core Minimal API with Microsoft.Identity.Web
Make an ASP.NET Core minimal API accept only access tokens that Microsoft Entra ID issued for it, with the scope you require: AddMicrosoftIdentityWebApi, a RequireScope policy and RequireAuthorization. Real 401, 200 and 403 answers from .NET 10 and Microsoft.Identity.Web 4.18.0.
15 min readHow to Get an Access Token for Your Own API with the Azure CLI
Test an API protected by Microsoft Entra ID from the command line: az account get-access-token --scope api://<client ID>/access_as_user, every field of its answer, keeping only the token, sending it with curl, and the AADSTS9002313 and 401 errors you can meet.
10 min readHow to Pre-Authorize a Client Application for Your API in Microsoft Entra ID
Let your API trust the apps you choose, so their users are never asked to consent: authorize the Azure CLI and your own single-page app for a scope on the Expose an API page, or with az rest and Microsoft Graph. Includes the AADSTS9002313 it prevents.
12 min readHow to Expose an API Scope in Microsoft Entra ID
Give your web API's app registration an Application ID URI and a scope such as access_as_user, so apps can ask Microsoft Entra ID for tokens to call it. Every field of the Expose an API page explained, plus the Azure CLI and Microsoft Graph route.
12 min readHow to Register a Web API in Microsoft Entra ID
Register your web API in Microsoft Entra ID so it can accept access tokens: every screen of the Microsoft Entra admin center, the IDs to copy, what the registration creates, and the same steps with az ad app create and az ad sp create.
12 min readFixed: Next.js Caches the MSAL Redirect Page with "Cache-Control: s-maxage=31536000"
MSAL Browser v5 says its redirect bridge page must be served with Cache-Control: no-store, but the production build of a Next.js 16 app served /redirect with Cache-Control: s-maxage=31536000 because the page is static. One headers() rule in next.config.ts fixed it.
10 min readFixed: "AADSTS9002313: Invalid request. Request is malformed or invalid." When the Azure CLI Gets a Token for Your Own API
az account get-access-token failed with AADSTS9002313 for the scope of a newly registered API. The --debug log showed AADSTS65001: nobody had consented to the Microsoft Azure CLI using that scope. Adding the Azure CLI as an authorized client application of the API fixed it.
11 min readMLflow Tutorial for Machine Learning in Python — Part 3: Log Artifacts and Datasets
Record the data every MLflow run trained, validated and tested on with mlflow.data and log_input, and save the run's files with it: a feature importance table with log_table, the model's full settings with log_dict and a CSV of every test prediction with log_artifact. Then compare datasets and files across runs in the MLflow UI. Real output from MLflow 3.18.0.
30 min readHow to Read a JSON or Text Artifact from MLflow in Python
Read a file a run logged straight into Python without saving it: mlflow.artifacts.load_dict for JSON and YAML, load_text for any text file, with runs:/ addresses. Real output from MLflow 3.18.0, and the errors for a wrong path or a file that isn't JSON.
5 min readHow to Download Artifacts from an MLflow Run
Copy the files an MLflow run logged to your computer: list them with mlflow artifacts list, download a file or a whole folder with mlflow.artifacts.download_artifacts in Python or mlflow artifacts download in a terminal, or use the UI's download button. Real output from MLflow 3.18.0.
6 min readHow to Compare a Logged Table Across Runs in the MLflow UI
Put a table logged with mlflow.log_table side by side for several runs in the MLflow UI's Artifact evaluation view: pick the table, the column that matches rows and the column to compare, and filter the rows. Real screens from MLflow 3.18.0, and why some runs don't appear.
5 min readHow to View a Run's Artifacts in the MLflow UI
Find, preview and download the files an MLflow run logged: the Artifacts tab's folder tree, previews for CSV, JSON, YAML and Markdown, the table view for log_table files, each file's path and download button, and side-by-side files on the compare page. Real screens from MLflow 3.18.0.
6 min readHow to Log a Folder of Files to MLflow with log_artifacts
Upload a whole folder to an MLflow run with mlflow.log_artifacts, and why artifact_path matters: without it, the folder's files land at the top level and the folder's name is lost. Real output and screens from MLflow 3.18.0, plus the silent result of a misspelled folder.
5 min readHow to Log a File to MLflow with log_artifact
Attach a file from your disk to an MLflow run with mlflow.log_artifact, at the top level or inside a folder with artifact_path, and see it in the MLflow UI with its preview and download button. Real output and screens from MLflow 3.18.0.
5 min readHow to Log Text to a File in MLflow with log_text
Save a string with an MLflow run as a file with mlflow.log_text, without writing it to disk first: a Markdown note that the MLflow UI renders, a plain text summary, and what happens with a second call or a list instead of a string. Real output and screens from MLflow 3.18.0.
5 min readHow to Log a Dictionary to MLflow as a JSON or YAML File
Save a Python dictionary with an MLflow run as a JSON or YAML file with mlflow.log_dict: how the file name picks the format, how JSON keeps your key order while YAML sorts it, and what happens to values JSON can't store. Real output and screens from MLflow 3.18.0.
6 min readHow to Load a Logged Table from MLflow with load_table
Read a table logged with mlflow.log_table back into pandas with mlflow.load_table: from every run of an experiment at once, labelled with each run's name, or from runs you pick by ID. Real output from MLflow 3.18.0, and the errors when no run has the table.
5 min readHow to Log a Table in MLflow with log_table
Save a pandas DataFrame or a dictionary with an MLflow run as a table with mlflow.log_table: the JSON file it writes, how the UI shows it as a table, why a second call adds rows instead of replacing them, and the file names it accepts. Real output from MLflow 3.18.0.
6 min readHow to Make an MLflow Dataset Digest Cover Every Row
MLflow fingerprints a pandas dataset from its first 10,000 rows and only its text and number columns, so a changed later row or date goes unnoticed. See it happen, see why in MLflow's source, and pass your own digest of every row and column to mlflow.data.from_pandas. Real output from MLflow 3.18.0.
8 min readHow to Check Whether Two MLflow Runs Used the Same Data
Compare the dataset digests of two MLflow runs to know whether they trained on the same data: in the runs table, with the Datasets filter, and in Python with mlflow.get_run. Real screens and output from MLflow 3.18.0, and what a matching digest does and doesn't prove.
7 min readHow to See Which Datasets an MLflow Run Used
Find the datasets an MLflow run logged with log_input, in four places: the runs table's Dataset column, the run page, the Data details panel with digest, source, schema and row count, and from the command line or Python. Real screens and output from MLflow 3.18.0.
8 min readHow to Log a Dataset in MLflow
Record which data an MLflow run used: turn a pandas DataFrame into an MLflow dataset with mlflow.data.from_pandas, log it with mlflow.log_input and a context such as training, and see its name, digest, source, schema and row count in the UI. Real output from MLflow 3.18.0.
8 min readHow to See Every Setting of a scikit-learn Model with get_params
List all of a scikit-learn model's settings, the ones you chose and every default, with get_params(): a dictionary of 18 settings for a RandomForestRegressor, how to read one, and why it shows settings, not what the model learned. Real output from scikit-learn 1.9.1.
5 min readHow to Get Feature Importances from a Random Forest in scikit-learn
Read which inputs a trained random forest relies on with feature_importances_ and feature_names_in_, put them in a sorted pandas table, and know what the numbers mean and where scikit-learn warns they can mislead. Real output from scikit-learn 1.9.1 on the California Housing data.
6 min readHow to Save a pandas DataFrame to a CSV File
Write a pandas DataFrame to a CSV file with to_csv, and why index=False matters: without it, pandas adds a nameless first column of row numbers that comes back as Unnamed: 0. Real files from pandas 3.1.1, and the error for a folder that doesn't exist.
4 min readHow to Sort a pandas DataFrame by a Column
Sort a pandas DataFrame by one column with sort_values: smallest first, largest first, by size whatever the sign with key=abs, and renumbered rows with ignore_index=True. Real output from pandas 3.1.1, and why the original table doesn't change.
5 min readHow to Join a pandas Series to a DataFrame by Index
Add a pandas Series to a DataFrame as a new column with DataFrame.join, matched by row label rather than position: what happens when the order differs or a label is missing, and the errors for an unnamed Series or a column name that's already taken. Real output from pandas 3.1.1.
5 min readHow to Create a pandas DataFrame from a Dictionary
Build a pandas DataFrame from a Python dictionary, one key per column: how pandas picks each column's type, why text columns show the str type in pandas 3, and the errors for lists of different lengths or single values. Real output from pandas 3.1.1.
5 min readHow to Build File Paths in Python with pathlib
Join folders and file names with pathlib's / operator instead of string concatenation, then read a path's name, suffix, stem and parent, and start from your home folder with Path.home(). Real output from Python 3.15.2, and the two mistakes that break a path.
5 min read