pandexai 0.1.0 → 0.1.2

package/.claude/commands/pandex.md

@@ -0,0 +1,28 @@
+---
+description: Run PandexAI commands (scan, gather, analyze) for deterministic data profiling
+---
+
+# Pandex
+
+You are handling a `/pandex` command. The arguments passed are: $ARGUMENTS
+
+Parse the first word as the subcommand, and everything after it as the filename (if any).
+
+## If subcommand is "scan"
+
+1. Read SKILL.md and commands/scan.md in this project for full instructions.
+2. If no filename was given, tell the user a filename is required for now (`/pandex gather` is planned but not built yet).
+3. Run: python scripts/profile.py <filename>
+   Use the project's virtual environment Python, not the system one:
+   - Windows: .pandex\venv\Scripts\python.exe
+   - Mac/Linux: .pandex/venv/bin/python
+4. The script prints a JSON object. Do NOT recompute, estimate, or guess any of these numbers yourself.
+5. Present the JSON results to the user as a clear, readable summary, and flag anything that looks like a data quality issue (high null %, suspicious types, low-cardinality columns that might be categorical).
+
+## If subcommand is "gather" or "analyze"
+
+Tell the user this command is planned but not built yet, and suggest `/pandex scan <file>` instead.
+
+## If no subcommand is recognized
+
+Show usage: `/pandex scan <file>` (gather and analyze coming soon).

package/README.md

@@ -1,3 +1,29 @@
-# PandexAI
-
-AI-native data profiling CLI. Under active development — not yet ready for use.
+# PandexAI
+
+AI-native data profiling CLI. PandexAI runs inside AI coding CLIs (Claude Code, Cursor, etc.) and handles the deterministic parts of data analysis - profiling columns, checking nulls, catching type issues - as real Python/pandas code, not AI guesses. The AI only handles judgment: interpreting results, flagging issues, and deciding what matters.
+
+## Why
+
+When working with an AI CLI on messy data, a lot of the groundwork does not need an LLM at all - it just needs pandas doing pandas things. PandexAI keeps that part deterministic and trustworthy, and lets the AI focus on the part that actually needs reasoning.
+
+## Install
+
+    npx pandexai init
+
+This creates an isolated Python environment inside your project (`.pandex/venv`), installs pandas into it, and adds the files PandexAI needs (`SKILL.md`, `commands/`, `scripts/`).
+
+## Usage
+
+Inside your AI CLI session, ask it to scan a file:
+
+    Please read SKILL.md and commands/scan.md, then run the scan command on yourfile.csv
+
+The AI will run the real profiling script and give you column types, null percentages, unique counts, and stats - plus its own read on anything that looks off.
+
+## Status
+
+Early and actively developed. Currently supports one command: `scan`. More coming.
+
+## License
+
+MIT

package/bin/pandex.js

@@ -57,7 +57,7 @@ if (installResult.status !== 0) {
 }
 
 const packageRoot = path.join(__dirname, "..");
-const filesToCopy = ["SKILL.md", "commands", "scripts"];
+const filesToCopy = ["SKILL.md", "commands", "scripts", ".claude"];
 
 function copyRecursive(src, dest) {
   const stat = fs.statSync(src);
@@ -81,4 +81,4 @@ for (const item of filesToCopy) {
 }
 
 console.log("");
-console.log("PandexAI is ready. Try: /pandex scan <your-file.csv> inside your AI CLI.");
+console.log("PandexAI is ready. Try: /pandex scan <your-file.csv> inside your AI CLI.");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pandexai",
-  "version": "0.1.0",
+  "version": "0.1.2",
   "description": "PandexAI - AI-native data profiling CLI. Run pandex init inside your project to get started.",
   "license": "MIT",
   "repository": {
@@ -12,9 +12,10 @@
     "pandex": "./bin/pandex.js"
   },
   "files": [
-    "bin",
-    "SKILL.md",
-    "commands",
-    "scripts"
-  ]
+  "bin",
+  "SKILL.md",
+  "commands",
+  "scripts",
+  ".claude"
+]
 }