Zenalyze

Zenalyze Public API Usage Guide

This section documents the public APIs exposed at the package root. These cover data loading, orchestration, and LLM-powered analytics workflows.

---

1. Data Backends

• PandasData

  Represents a single pandas DataFrame plus metadata.

  Usually created internally by PandasDataLoad, but you can also create and pass these directly into Zenalyze.

• PandasDataLoad(data_location: str)

  Loads all CSV/Excel files from a directory as pandas DataFrames and wraps each in a PandasData object.

  Usage:

    loader = PandasDataLoad("./data")
    datasets = loader.data
  

• SparkData

  Wraps a Spark DataFrame with metadata.

  Also passable directly into Zenalyze if you already constructed your own Spark DataFrames.

• SparkDataLoad(spark_session, data_location: str)

  Loads CSV/Excel/Parquet files using Spark and returns Spark-backed Data objects.

  Usage:

    spark = SparkSession.builder.getOrCreate()
    loader = SparkDataLoad(spark, "./data")
    datasets = loader.data
  

---

2. High-Level Orchestrators

• Zenalyze

  The main orchestration engine that:

  • loads provided datasets into the caller’s globals()
  • builds prompts using metadata + chat history
  • generates Python code from natural-language queries
  • executes code with full context
  • provides a built-in buddy assistant (zen.buddy(“…”))
  • maintains and summarizes long histories internally

  ✔️ Accepts ANY mix of either of following:

  • PandasData and/or PandasDataLoad
  • SparkData and/or SparkDataLoad

  And as many as you want via *args.

  Example:

    from zenalyze import Zenalyze, PandasDataLoad, PandasData
      
    loader = PandasDataLoad("./data")
    custom_table = PandasData(df=my_df, name="mytable")
      
    zen = Zenalyze(globals(), loader, custom_table)
      
    zen.do("calculate total revenue")
    zen.buddy("What have we done so far?")
  

  This gives maximum flexibility — you can load some tables from disk and hand-craft others dynamically.

• TestZen

  A mock/testing version of Zenalyze with:

  • no real LLM calls
  • predictable dummy responses
  • fully functional prompt logic

  ✔️ Also accepts any combination of Data/ DataLoad objects:

    from zenalyze import TestZen, PandasDataLoad
      
    loader = PandasDataLoad("./data")
    zen = TestZen(globals(), loader)
      
    zen.do("summary stats")
    zen.buddy("What next?")
  

---

3. Convenience Constructors

• create_zenalyze_object_with_env_var_and_last5_hist

  A quick way to create a fully configured Zenalyze instance:

  • backend auto-detected (pandas or spark)
  • loads data from a directory
  • sets history_retention=5
  • sets return_query=False for clean .do() usage

  Usage:

    zen = create_zenalyze_object_with_env_var_and_last5_hist(
        globals(),
        "./data"
    )
  

• create_testzen_object_with_env_var_and_last5_hist

  Same as above, but returns a TestZen instance.

---

4. Chat Helpers

• CodeSummarizerLLM

  Internal summarizer used by Zenalyze to compress long histories. Still available for direct use if needed.

  Usage:

    summ = CodeSummarizerLLM()
    summary = summ.summarize(raw_history_text)
  

• BuddyLLM

  The assistant that powers:

    zen.buddy("my question")
  

  You rarely need to instantiate this manually because Zenalyze already contains:

    zen.buddy_llm
  

  But optional standalone use is allowed.

---

✔️ Final Summary

API	Accepts What	Purpose
PandasData	Individual pandas DataFrames	Manual creation of Data objects
PandasDataLoad	Directory path	Load multiple pandas tables + metadata
SparkData	Spark DataFrames	Manual creation of Spark Data objects
SparkDataLoad	SparkSession + directory	Load Spark-backed tables + metadata
Zenalyze	Any mix of either Pandas or Spark DataLoad + Data objects (*args)	Main LLM analysis engine
TestZen	Any mix of either Pandas or Spark DataLoad + Data objects (*args)	Mock testing engine
create_zenalyze_object_with_env_var_and_last5_hist	Directory, SparkSession only for Spark usage	Quick Zenalyze setup
create_testzen_object_with_env_var_and_last5_hist	Directory, SparkSession only for Spark usage	Quick TestZen setup
CodeSummarizerLLM	Raw history	Summarization utility
BuddyLLM	Summary context	Q&A assistant

This site is open source. Improve this page.