duyanta123/dsh-data-insight：DSH 数据洞察技能插件
`duyanta123/dsh-data-insight` 是一款基于 DSH 生态的数据洞察技能插件，旨在将 CSV、粘贴表格、SQL 结果或 DuckDB 数据转化为包含业务结论、指标和图表的结构化 Markdown 报告。该插件采用纯指令型实现，零依赖、零构建，通过 LLM 驱动宿主工具完成计算。其核心功能涵盖五种数据源支持及“输入受理、数据探查、指标计算、图表呈现、报告产出”的五阶段流水线，提供汇总统计、同环比、TopN 及异常值检测等指标，并支持 Markdown、Mermaid 及 ASCII 三种图表呈现方式。用户可通过 `dsh plugin` 命令安装并在会话中直接触发分析。
Read Moredsh-data-agent: Enabling DeepSeek Harness to Connect to Databases for Analysis via Dialogue
dsh-data-agent is a DSH data analysis plugin maintained by omdsh-dev, supporting common databases such as MySQL, PostgreSQL, and ClickHouse. After installation, users can select "Data Mode" to ask questions in natural language; DSH will automatically generate SQL, execute queries, and provide business insights. It also offers Catalog data governance and HTML analysis reports. This article verifies installation commands, typical usage, and security precautions based on the SkillHub directory page and the GitHub README.
Read MorePandas Tutorial for Beginners: Missing Value Handling from Entry to Practice
This article introduces methods for handling missing values in data analysis. Missing values refer to non-valid values in a dataset, represented as `NaN` in pandas. Before processing, it is necessary to first check: `isnull()` to mark missing values, `isnull().sum()` to count the number of missing values in each column, and `info()` to view the overall distribution of missing values. Processing strategies are divided into deletion and imputation: Deletion uses `dropna()`, which deletes records containing missing values by row (default) or by column; Imputation uses `fillna()`, including fixed values (e.g., 0), statistical measures (mean/median for numerical values, mode for categorical values), and forward/backward filling (`ffill/bfill`, suitable for time series). Taking e-commerce order data as an example, the case first checks for missing values, then uses the mean to impute the "amount" column and the mode to impute the "payment method" column. The core steps of processing are: check for missing values → select a strategy (delete for extremely few values, impute for many values or key data) → verify the result. It is necessary to flexibly choose methods based on the characteristics of the data.
Read MoreLearning pandas from Scratch: A Step-by-Step Guide to Reading CSV Files
This article introduces the introductory steps to learning pandas for data processing, with the core being reading CSV files and performing basic data operations. First, pandas is likened to the "steward" of data processing, and reading CSV is the first step in data analysis. The steps include: installing pandas (using `pip install`, or skipping if pre-installed with Anaconda/Jupyter); importing pandas as `import pandas as pd`; reading the CSV file with `pd.read_csv()` to generate a DataFrame; viewing data using `head()`/`tail()` for preview, `info()` to check data types and missing values, and `describe()` for numerical statistics; handling special formats such as Chinese garbled characters (via `encoding`), delimiters (via `sep`), and no header rows (via `names`). The article concludes by summarizing the basic skills acquired, noting that this is just the beginning of data processing, and subsequent advanced operations like filtering and cleaning can be learned next.
Read More