Analyze data through
conversation.

Upload tabular datasets, ask questions in natural language, and receive profiles, charts, dashboards, and AI-assisted analysis — no SQL required.

Try it free See how it works
mnemos.ai demo
Built on

Best-in-class open source, end-to-end.

Every layer of the stack is production-grade, open, and replaceable — no proprietary lock-in.

FastAPI
Backend API
Async Python API framework powering all data ingestion, analysis routing, and artifact serving endpoints.
LangGraph
Agent Orchestration
Structured agentic workflow with clarification gates, validated plan nodes, and deterministic execution — not a prompt loop.
React 19
Frontend
Conversational workspace built with React's latest features — streaming responses, artifact panels, and persistent chat history.
pandas
Data Processing
Handles CSV, XLSX, Parquet, and JSON ingestion with schema detection, type inference, and quality reporting.
Matplotlib
Visualization
Deterministic chart rendering — every bar chart, line graph, and histogram is validated and stored as a reproducible artifact.
OpenAI
LLM Integration
Structured-output LLM calls for intent parsing and plan generation. Works with heuristic fallback when no API key is configured.
Features

Everything you need
from upload to insight.

A structured pipeline with validation, deterministic rendering, and persistent storage — not just a prompt wrapper around an LLM.

Data Ingestion

Upload any tabular dataset and get instant profiling.

Drop a CSV, XLSX, Parquet, or JSON file. The pandas-based quality pipeline runs automatically — schema detection, missing-value analysis, type inference, and a quality report, all without configuration.

View on GitHub
Datasets Quality Report
ColumnTypeMissingStatus
revenuefloat640%Clean
regionobject0%Clean
datedatetime2%2 missing
categoryobject0%Clean
units_soldint640%Clean
AI Workflow

Ask any question. The agent does the rest.

Powered by LangGraph, the analysis agent interprets your intent, builds a validated plan, and asks for clarification when a request is ambiguous — no silent hallucination, no guessing.

Explore the workflow
Explore sales_q3.csv
M
Dataset loaded. 1,240 rows, 8 columns. No critical issues detected. What would you like to analyze?
U
Show monthly revenue trends for Q3
M
Analyzing revenue by month for Q3…
M
Visualization

Deterministic charts — no hallucinated visualizations.

Every chart is rendered via Matplotlib with validated specifications. Bar, line, histogram, and scatter plots — fully reproducible, stored as artifacts, and served directly through the FastAPI /storage route.

View renderers
Chart Output revenue_q3_bar.png
MONTHLY REVENUE — Q3 2024
Jul
Aug
—
Sep
—
Oct
—
Nov
Dec
Dashboard

Multi-panel dashboards from a single request.

Ask for a full dashboard and receive multiple chart panels assembled as a single artifact. The React workspace displays all panels inline in the chat, stored permanently in the /storage route.

Explore workspace
Dashboard 4 panels
TOTAL REVENUE
$2.4M
↑ 12.4% vs last quarter
UNITS SOLD
18,420
↑ 8.1% vs last quarter
TOP REGION
West
34% of total revenue
AVG ORDER
$130
↓ 2.3% vs last quarter
How It Works

From file to insight in three steps.

No configuration, no SQL, no dashboarding tool. Just upload, ask, and analyze.

01
Upload your dataset
Drop a CSV, XLSX, Parquet, or JSON file. The quality pipeline runs instantly — schema detection, missing-value analysis, and a quality report, all automatic.
02
Ask a question
Type your question in plain language — "Show revenue trends for Q3" or "Which category has the most missing values?" The LangGraph agent interprets intent and builds a validated plan.
03
Get your analysis
Receive text answers, bar charts, line graphs, histograms, scatter plots, or full multi-panel dashboards — all rendered deterministically and persisted in the chat history.

Ready to analyze your data
through conversation?

Open source, self-hosted, and free. Clone the repository and start asking questions in minutes.

View on GitHub See how it works