AtlasMind

Talk to your data. Get production-ready insights in <200 ms.

Live Demo

---

Why AtlasMind

AtlasMind is a conversational BI platform for MongoDB that combines:

• Natural language analytics
• Voice-to-insight pipelines
• Read-only safety enforcement
• Fast chart-first exploration

You ask a question in plain English. AtlasMind profiles schema context, generates aggregation logic, validates it for safety, executes the query, and returns a visual answer with explanation.

---

Tech Stack (Shields)

Frontend

Backend

AI and Speech

DevOps and CI

---

What Makes It Different

• Schema-aware prompting: generated pipelines are grounded in profiled collection/field metadata.
• Safety-first query execution: blocked stages/operators are rejected before execution.
• Session-based security: httpOnly JWT cookie for protected analytics routes.
• Voice-to-query support: upload audio and get chartable results in one flow.
• Persistent dashboard pins: save and refresh important query snapshots.

---

Architecture Snapshot

flowchart TD
    subgraph CLIENT["🖥️  Client Layer"]
        direction TB
        U("👤 User")
        UI("⚡ React + Vite UI")
        VOICE("🎙️ Voice Upload")
        PIN("📌 Pin to Dashboard")
    end

    subgraph GATEWAY["🔐  API Gateway  ·  Express + JWT"]
        direction TB
        AUTH("🔑 Auth · httpOnly Cookie")
        ROUTE("🔀 Route · /query  /voice  /pin")
        SCHEMA("🗂️ Schema Profiler · Field & Collection Metadata")
        FEWSHOT("🧠 Few-Shot Retriever · Cosine Similarity")
    end

    subgraph AI["🤖  AI Pipeline  ·  Groq LPU"]
        direction TB
        LLM("⚡ LLM · gpt-oss-120b / compound / qwen3.8")
        GUARD("🛡️ Safety Guard · Read-Only Enforcer")
        WHISPER("🎧 Whisper · whisper-large-v3-turbo")
    end

    subgraph DATA["🗄️  Data Layer  ·  MongoDB"]
        direction TB
        AGG("⚙️ Aggregation Executor")
        RESULT("📊 Charts + Explanation")
        DASH("🗃️ Dashboard Store")
    end

    U -- "Text Query" --> UI
    U -- "Audio File" --> VOICE
    UI --> AUTH
    VOICE --> AUTH
    AUTH --> ROUTE

    ROUTE --> SCHEMA
    SCHEMA --> FEWSHOT
    FEWSHOT --> LLM

    VOICE -- "Buffer" --> WHISPER
    WHISPER -- "Transcript" --> LLM

    LLM --> GUARD
    GUARD -- "✅ Safe MQL" --> AGG
    GUARD -- "🚫 Blocked" --> ROUTE

    AGG --> RESULT
    RESULT --> UI
    RESULT --> PIN
    PIN --> DASH

    classDef clientNode  fill:#1e293b,stroke:#6366f1,color:#e2e8f0,rx:8
    classDef gatewayNode fill:#1e293b,stroke:#0ea5e9,color:#e2e8f0,rx:8
    classDef aiNode      fill:#1e293b,stroke:#f59e0b,color:#e2e8f0,rx:8
    classDef dataNode    fill:#1e293b,stroke:#22c55e,color:#e2e8f0,rx:8

    class U,UI,VOICE,PIN clientNode
    class AUTH,ROUTE,SCHEMA,FEWSHOT gatewayNode
    class LLM,GUARD,WHISPER aiNode
    class AGG,RESULT,DASH dataNode

Loading

Documentation Portal

Guide	Purpose
Architecture	System design, flows, and technical patterns
Setup	Local installation and configuration
Features	Product capabilities and usage ideas
API	Endpoint contracts and response shapes
Development	Workflow, tests, and conventions
Deployment	Netlify + Render deployment path
Troubleshooting	Common failures and fixes

Model Roster

AtlasMind supports dynamic model switching from the chat UI. The active Groq model lineup is:

Model	Provider	Role	Speed
openai/gpt-oss-120b	OpenAI Community	Default — flagship precision MQL generation & schema reasoning	Balanced
openai/gpt-oss-20b	OpenAI Community	Lightweight ultra-fast MQL prototyping	Fastest
groq/compound	Groq Labs	Multi-model agentic router with safety validation	Sub-150ms
qwen/qwen3.8-27b	Alibaba Cloud	Frontier coder architecture for complex nested Mongo queries	Fast
qwen/qwen3.6-27b	Alibaba Cloud	Structured data reasoning for relational joins and aggregations	Fast
whisper-large-v3-turbo	OpenAI	Voice transcription (audio → text)	Realtime

---

Author

Lakshman Bhukya
Full-Stack Developer and AI Enthusiast

---