AI-Assisted Collaborative Analytics Platform
A human-AI collaborative analytics workspace where an LLM co-pilot assists analysts with EDA, chart generation, and insight narration via tool use, streaming responses, and multi-turn context — delivering 3× faster analysis cycles.
Overview
Key Features
Smart EDA: AI auto-scans the dataset for missing values, outliers, and distribution characteristics, generating a descriptive statistics summary
Collaborative code generation: analyst describes needs in natural language; AI produces executable Python (pandas/seaborn) with one-click run
Dynamic chart generation: LLM tool calls auto-select chart type and configure ECharts parameters based on data semantics
Insight narration: after each chart, AI writes a business-language interpretation covering trends, anomalies, and action recommendations
Streaming output: LLM responses pushed via SSE stream so analysts see the AI reasoning in real time, reducing perceived latency
Methodology
Backend: FastAPI + LangChain. Frontend: Next.js + ECharts. The AI co-pilot has 5 tool types: data summary, statistical computation, chart config generation, Python execution sandbox, and insight writing. Tool call results return structured JSON; the frontend renders different UI components (code block, chart, text card) based on tool type. Conversation history is summarized and appended to the system prompt to maintain long-session coherence.
Tech Stack
Project Info
AI Cowork Analytics Demo
Pick an analysis query and watch the LLM copilot drive data insights through tool calls and streaming output.
AI Co-work Data Analysis Platform
Select an analysis task — the LLM co-pilot calls data tools, streams insights, and renders inline charts.
Select Analysis Task