[kit] [mission-possible]Add EDA Copilot — AI-powered Exploratory Data Analysis agent [agentkit-challenge]

kits/agentic/eda-copilot/.env.example

@@ -0,0 +1,14 @@
+# Lamatic API Configuration
+LAMATIC_API_URL="https://your-org-your-project.lamatic.dev"
+LAMATIC_PROJECT_ID="your-project-id"
+LAMATIC_API_KEY="your-api-key"
+
+# EDA Copilot Flow IDs
+# Step 1: Analyzes column schema, data types, and data quality
+EDA_SCHEMA_ANALYSIS_FLOW_ID="your-schema-analysis-flow-id"
+
+# Step 2: Generates statistical insights and distribution analysis
+EDA_STATISTICAL_INSIGHTS_FLOW_ID="your-statistical-insights-flow-id"
+
+# Step 3: Produces ML readiness score and preprocessing recommendations
+EDA_ML_READINESS_FLOW_ID="your-ml-readiness-flow-id"

kits/agentic/eda-copilot/.gitignore

@@ -0,0 +1,34 @@
+# dependencies
+/node_modules
+/.pnp
+.pnp.js
+
+# testing
+/coverage
+
+# next.js
+/.next/
+/out/
+
+# production
+/build
+
+# misc
+.DS_Store
+*.pem
+
+# debug
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+
+# local env files
+.env*.local
+.env
+
+# vercel
+.vercel
+
+# typescript
+*.tsbuildinfo
+next-env.d.ts

kits/agentic/eda-copilot/README.md

@@ -0,0 +1,176 @@
+# EDA Copilot — AI-Powered Exploratory Data Analysis
+
+> Upload any CSV. Get instant AI-powered statistical insights, correlation analysis, outlier detection, and ML readiness assessment — all powered by [Lamatic.ai](https://lamatic.ai) flows.
+
+[![Deploy with Vercel](https://vercel.com/button)](https://vercel.com/new/clone?repository-url=https://github.com/Lamatic/AgentKit&root-directory=kits/agentic/eda-copilot)
+
+---
+
+## What It Does
+
+EDA Copilot is a data scientist's shortcut from raw CSV to actionable insights. Instead of writing pandas code, you drop a file and a three-stage Lamatic.ai agent pipeline does the work:
+
+| Stage | Flow | What it does |
+|-------|------|--------------|
+| 1 | Schema Analysis | Detects column types, missing data, and data quality issues |
+| 2 | Statistical Insights | Interprets distributions, correlations, and outliers |
+| 3 | ML Readiness | Scores dataset and generates preprocessing recommendations |
+
+The client-side engine computes raw statistics (mean, std, quartiles, correlation matrix, skewness) and sends a compact summary payload to Lamatic — keeping API costs low even for large files.
+
+---
+
+## Demo
+
+![EDA Copilot Screenshot](public/preview.png)
+
+Try it live → **[agent-kit-eda-copilot.vercel.app](https://agent-kit-eda-copilot.vercel.app)**
+
+---
+
+## Quick Start
+
+### 1. Clone and install
+
+```bash
+git clone https://github.com/Lamatic/AgentKit.git
+cd AgentKit/kits/agentic/eda-copilot
+npm install
+```
+
+### 2. Set up Lamatic flows
+
+1. Sign up at [lamatic.ai](https://lamatic.ai)
+2. Import each flow from the `/flows` directory into your Lamatic project
+3. Copy each flow's ID into `.env.local`
+
+```bash
+cp .env.example .env.local
+```
+
+```env
+LAMATIC_API_URL=https://your-project.lamatic.ai/api/v1
+LAMATIC_PROJECT_ID=your-project-id
+LAMATIC_API_KEY=your-api-key
+
+EDA_SCHEMA_ANALYSIS_FLOW_ID=your-schema-analysis-flow-id
+EDA_STATISTICAL_INSIGHTS_FLOW_ID=your-statistical-insights-flow-id
+EDA_ML_READINESS_FLOW_ID=your-ml-readiness-flow-id
+```
+
+### 3. Run
+
+```bash
+npm run dev
+```
+
+Open [http://localhost:3000](http://localhost:3000) and drop a CSV file to start analyzing.
+
+---
+
+## Lamatic Flow Setup
+
+The `/flows` directory contains three JSON files describing each flow's logic. To set them up in Lamatic Studio:
+
+1. **Go to your Lamatic project → Flows → New Flow**
+2. Use the `systemPrompt` and `userPromptTemplate` from each flow JSON as your LLM node configuration
+3. Set the input schema to match the `inputs` field in each flow JSON
+4. Deploy the flow and copy its Flow ID to `.env.local`
+
+### Flow Prompts at a Glance
+
+**Flow 1 — Schema Analysis** (`eda-schema-analysis.json`)
+- Input: `datasetSummary` (object), `fileName` (string)
+- Output: Markdown schema report with data quality score
+
+**Flow 2 — Statistical Insights** (`eda-statistical-insights.json`)
+- Input: `datasetSummary` (object), `schemaInsights` (string from Flow 1)
+- Output: Markdown insights on distributions, correlations, outliers
+
+**Flow 3 — ML Readiness** (`eda-ml-readiness.json`)
+- Input: `datasetSummary`, `schemaInsights`, `statisticalInsights` (from Flows 1 & 2)
+- Output: ML readiness score 0–100 + preprocessing checklist
+
+---
+
+## Architecture
+
+```
+┌─────────────────────────────────────────────┐
+│               Browser (Next.js)              │
+│                                              │
+│  CSV Upload → PapaParse → computeSummary()  │
+│        ↓ (compact stats payload)            │
+│  /api/analyze?step=schema                   │
+│       ↓                                     │
+│  /api/analyze?step=statistical              │
+│       ↓                                     │
+│  /api/analyze?step=mlReadiness              │
+│       ↓                                     │
+│  Results rendered with Recharts + Markdown  │
+└──────────────┬──────────────────────────────┘
+               │ REST API calls
+┌──────────────▼──────────────────────────────┐
+│            Lamatic.ai Flows                  │
+│                                              │
+│  Flow 1: Schema Analysis (LLM node)         │
+│  Flow 2: Statistical Insights (LLM node)    │
+│  Flow 3: ML Readiness (LLM node)            │
+└─────────────────────────────────────────────┘
+```
+
+**Key design decision:** All heavy computation (statistics, correlations, skewness) runs client-side in the browser using pure TypeScript. Only a compact JSON summary is sent to Lamatic flows — this keeps token usage low and makes the app fast even for 100k+ row datasets.
+
+---
+
+## Tech Stack
+
+- **Framework**: Next.js 15 (App Router)
+- **Styling**: Tailwind CSS
+- **CSV Parsing**: PapaParse (client-side, zero server load)
+- **Charts**: Recharts
+- **Markdown**: react-markdown
+- **AI Orchestration**: Lamatic.ai (3 sequential LLM flows)
+
+---
+
+## One-Click Deploy
+
+[![Deploy with Vercel](https://vercel.com/button)](https://vercel.com/new/clone?repository-url=https://github.com/Lamatic/AgentKit&root-directory=kits/agentic/eda-copilot)
+
+You'll be prompted to set the environment variables during deployment.
+
+---
+
+## Project Structure
+
+```
+eda-copilot/
+├── app/
+│   ├── api/analyze/route.ts   # Lamatic flow API gateway
+│   ├── layout.tsx
+│   ├── page.tsx               # Main UI
+│   └── globals.css
+├── actions/
+│   └── orchestrate.js         # Flow definitions & API config
+├── components/
+│   ├── FileUpload.tsx          # Drag-and-drop CSV uploader
+│   ├── DataPreview.tsx         # Column chips + sample table
+│   ├── AnalysisProgress.tsx    # Step-by-step progress tracker
+│   └── AnalysisResults.tsx     # Full results with charts
+├── lib/
+│   └── utils.ts               # Client-side EDA engine
+├── flows/
+│   ├── eda-schema-analysis.json
+│   ├── eda-statistical-insights.json
+│   └── eda-ml-readiness.json
+├── config.json                # Kit metadata
+├── .env.example
+└── README.md
+```
+
+---
+
+## License
+
+MIT — part of the [Lamatic AgentKit](https://github.com/Lamatic/AgentKit) open-source collection.

kits/agentic/eda-copilot/actions/orchestrate.js

@@ -0,0 +1,59 @@
+/**
+ * EDA Copilot — Lamatic SDK Configuration & Flow Definitions
+ *
+ * Uses the official `lamatic` npm SDK to call three sequential flows:
+ *  1. Schema Analysis      → column types, data quality score
+ *  2. Statistical Insights → distributions, correlations, outliers
+ *  3. ML Readiness         → preprocessing recommendations, readiness score
+ *
+ * Install: npm install lamatic
+ * Docs:    https://lamatic.ai/docs/api-integration/sdk
+ */
+
+import { Lamatic } from "lamatic";
+
+export const lamatic = new Lamatic({
+  endpoint:  process.env.LAMATIC_API_URL,
+  projectId: process.env.LAMATIC_PROJECT_ID,
+  apiKey:    process.env.LAMATIC_API_KEY,
+});
+
+/**
+ * Flow definitions — IDs come from .env after deploying each flow in Lamatic Studio.
+ * Each flow's input keys must match the payload you send via lamatic.executeFlow().
+ */
+export const flows = {
+  /**
+   * Flow 1 — Schema Analysis
+   * Lamatic input schema:  datasetSummary (Object), fileName (String)
+   * Lamatic output key:    generatedResponse (String — markdown)
+   */
+  schemaAnalysis: {
+    id: process.env.EDA_SCHEMA_ANALYSIS_FLOW_ID,
+    inputKeys: ["datasetSummary", "fileName"],
+  },
+
+  /**
+   * Flow 2 — Statistical Insights
+   * Lamatic input schema:  datasetSummary (Object), schemaInsights (String)
+   * Lamatic output key:    generatedResponse (String — markdown)
+   */
+  statisticalInsights: {
+    id: process.env.EDA_STATISTICAL_INSIGHTS_FLOW_ID,
+    inputKeys: ["datasetSummary", "schemaInsights"],
+    dependsOn: "schemaAnalysis",
+  },
+
+  /**
+   * Flow 3 — ML Readiness
+   * Lamatic input schema:  datasetSummary (Object), schemaInsights (String), statisticalInsights (String)
+   * Lamatic output key:    generatedResponse (String — markdown)
+   */
+  mlReadiness: {
+    id: process.env.EDA_ML_READINESS_FLOW_ID,
+    inputKeys: ["datasetSummary", "schemaInsights", "statisticalInsights"],
+    dependsOn: ["schemaAnalysis", "statisticalInsights"],
+  },
+};
+
+export default flows;

kits/agentic/eda-copilot/app/api/analyze/route.ts

@@ -0,0 +1,97 @@
+import { NextRequest, NextResponse } from "next/server";
+import { Lamatic } from "lamatic";
+
+// Lazy-init so the module can be imported at build time without real env vars
+function getLamatic() {
+  return new Lamatic({
+    endpoint: process.env.LAMATIC_API_URL!,
+    projectId: process.env.LAMATIC_PROJECT_ID!,
+    apiKey: process.env.LAMATIC_API_KEY!,
+  });
+}
+
+const FLOW_IDS = {
+  schema: process.env.EDA_SCHEMA_ANALYSIS_FLOW_ID,
+  statistical: process.env.EDA_STATISTICAL_INSIGHTS_FLOW_ID,
+  mlReadiness: process.env.EDA_ML_READINESS_FLOW_ID,
+};
+
+export async function POST(req: NextRequest) {
+  try {
+    const lamatic = getLamatic();
+    const body = await req.json();
+    const { datasetSummary, fileName, step } = body;
+
+    if (!datasetSummary) {
+      return NextResponse.json({ error: "datasetSummary is required" }, { status: 400 });
+    }
+
+    // ── Step 1: Schema Analysis ────────────────────────────────────────────
+    if (step === "schema") {
+      const response = await lamatic.executeFlow(FLOW_IDS.schema!, {
+        datasetSummary,
+        fileName: fileName || "dataset.csv",
+      });
+
+      if (response.status !== "success") {
+        throw new Error(response.message || "Schema analysis flow failed");
+      }
+
+      return NextResponse.json({ success: true, step: "schema", result: response.result });
+    }
+
+    // ── Step 2: Statistical Insights ──────────────────────────────────────
+    if (step === "statistical") {
+      const { schemaInsights } = body;
+      if (!schemaInsights) {
+        return NextResponse.json(
+          { error: "schemaInsights is required for step 2" },
+          { status: 400 }
+        );
+      }
+
+      const response = await lamatic.executeFlow(FLOW_IDS.statistical!, {
+        datasetSummary,
+        schemaInsights,
+      });
+
+      if (response.status !== "success") {
+        throw new Error(response.message || "Statistical insights flow failed");
+      }
+
+      return NextResponse.json({ success: true, step: "statistical", result: response.result });
+    }
+
+    // ── Step 3: ML Readiness ──────────────────────────────────────────────
+    if (step === "mlReadiness") {
+      const { schemaInsights, statisticalInsights } = body;
+      if (!schemaInsights || !statisticalInsights) {
+        return NextResponse.json(
+          { error: "schemaInsights and statisticalInsights required for step 3" },
+          { status: 400 }
+        );
+      }
+
+      const response = await lamatic.executeFlow(FLOW_IDS.mlReadiness!, {
+        datasetSummary,
+        schemaInsights,
+        statisticalInsights,
+      });
+
+      if (response.status !== "success") {
+        throw new Error(response.message || "ML readiness flow failed");
+      }
+
+      return NextResponse.json({ success: true, step: "mlReadiness", result: response.result });
+    }
+
+    return NextResponse.json(
+      { error: "Invalid step. Use: schema | statistical | mlReadiness" },
+      { status: 400 }
+    );
+  } catch (err) {
+    const message = err instanceof Error ? err.message : "Unknown error";
+    console.error("[EDA Copilot API Error]", message);
+    return NextResponse.json({ error: message }, { status: 500 });
+  }
+}