INTENT-AWARE NLP ROUTING

The Unbroken AI Experience

Heros automatically coordinates multiple specialized engines to keep your productivity running at 100% uptime. Explore our systems below.

View on GitHub
0
Failover Tiers
0%
Uptime Target
0ms
Triage Latency
0/7
Active Monitoring
3-TIER FALLBACK ENGINE — LIVE

Meet Baymax

Heros routes every request through Baymax, an intent-aware premium dispatcher that fails over across Gemini, OpenRouter, and Groq in milliseconds. One provider outage never means a broken conversation.

Encrypted keys pgvector RAG Django 5.2
Try Baymax
BAYMAX ROUTING ONLINE
Gemini
Primary reasoning model
tier 1
OpenRouter
6-model fallback pool
tier 2
Groq
Low-latency inference
tier 3
Outage Simulation Sandbox
[SYSTEM] Baymax coordinator initialized... Status: ONLINE
HALO ROUTING ENGINE ACTIVE
User Request Received
Halo Coordinator
Intent Routing & Fast Triage
Casual / Fast Query
Execute Instantly
Heavy Reasoning
Route to Baymax
GATEWAY COORDINATOR

Meet Halo

Halo is the foundational coordinator of the ecosystem. Halo triages request types, responds to casual conversations instantly, and prompts Baymax for high-level operations.

Intelligent routing Instant coordination Automatic triage
Try Halo
Your data, in a conversation

Explore Infinsight

Drop in a spreadsheet and ask questions the way you'd ask a colleague. Infinsight embeds every row, retrieves the relevant ones, and runs real Pandas computations behind the scenes.

  • Upload a CSV, Excel, or PDF file
  • Rows are embedded and stored in pgvector
  • Ask in plain English — get computed answers back
Open Infinsight
infinsight — sales_q2.xlsx
sales_q2.xlsx · 4,208 rows indexed
Which region underperformed in June?
The South region underperformed by 18% in June, driven mostly by a drop in repeat orders.
Zeno
Eco
Plus

Hello! I'm Zeno, your mini AI assistant.

Can you break down the python event loop?

Type your message...
Send
Heros, anywhere on the web

Meet Zeno

Your personal mini AI assistant browser extension. Get instant answers without switching tabs, powered by Baymax's resilient routing.

  • Ask Zeno Plus: Right-click highlighted text to analyze instantly
  • Voice Chat: Talk naturally with instant interruption support
  • Temporary Chats: Quick answers without cluttering your history
/ Download for Edge & Chrome
Zuno inside Zeno

Meet Zuno

Your intelligent built-in music assistant. Control YouTube and YouTube Music seamlessly through the Zeno extension using your voice or a dedicated mini-player.

  • Silent background control: No more switching tabs.
  • Voice commands: "Play [Song]", "Pause", "Next", "Skip Ad"
  • Music Radar: Instantly bring your hidden music tab to the front.
/ Download for Edge & Chrome
Album Art
I'm Scared
Anirudh Ravichander
1:12 -2:24
Advanced NLP Pipeline

Adaptive Personas & Fast Mode

Heros dynamically shifts its intellect, tone, and response budgets to match your goals, with ultra-fast latency options.

Adaptive Personas

The advanced NLP engine routes tasks to specific model personalities based on your intent.

  • Chat (ChatGPT style): Natural, friendly, and engaging conversation.
  • Coding (Claude style): Production-ready code blocks and design logic.
  • Search (Gemini/Grok style): Real-time synthesized search facts.
  • Voice (ChatGPT Voice style): Concise, natural spoken phrases.

Groq Fast Mode & Budgeting

Toggle Fast Mode to bypass heavy reasoning paths for rapid-fire responses in milliseconds.

  • Instant Replies: Accelerated fallback paths for quick Q&A.
  • Response Budgeting: Automatically adjusts length by intent.
  • Resource Efficiency: Snappy, clean dialogs without context bloat.
  • Halo Integration: Zero-login foundational coordination.
Core pillars

Four systems, one assistant.

Everything Heros does is built around one idea: never leave the user staring at an error.

Multi-Model Fallback Engine

Baymax detects intent and silently reroutes across Gemini, OpenRouter, and Groq the instant a provider degrades.

  • Intent-aware routing
  • Zero-downtime handoff
  • Per-user API keys

Infinsight RAG Analytics

Upload a CSV or Excel file and chat with it in plain English — no formulas, no SQL, no pivot tables.

  • pgvector semantic search
  • Pandas execution layer
  • Chart-ready answers

Voice, Web & File Analysis

Speak instead of typing, pull live answers from the web, or hand over a PDF, image, or document to parse.

  • Speech-to-text input
  • Real-time web search
  • OCR & document parsing

Unbroken Chat Context

Conversation memory and smart routing persist across every model switch, so the thread never resets.

  • Persistent session history
  • Cross-model context transfer
  • NLP-based intent detection
Failover order

How it fails gracefully.

A fixed, predictable chain — each tier only activates if the one before it can't respond.

TIER 1

Gemini

Primary brain for general reasoning, coding help, and conversation. Handles the majority of requests.

TIER 2

OpenRouter

A pool of six backup models. Baymax picks the best available one the moment Gemini is unreachable.

TIER 3

Groq

Ultra-low-latency inference as the final safety net, keeping responses fast even under load.

Tech Stack & Ecosystem

The engines behind Heros.

Built with modern, scalable, and resilient technologies to ensure your workflow never breaks.

AI Models (LLMs)

Powered by a robust triple-tier fallback system:

  • Gemini: Primary reasoning engine
  • OpenRouter: Secondary fallback pool
  • Groq: Ultra-fast inference

Data & Vector Storage

Seamlessly blending relational data with high-dimensional vector embeddings.

  • PostgreSQL: Core relational database
  • pgvector: Vector embedding store
  • RAG Analytics: Seamless context retrieval

Zeno Extension

Our companion browser extension that brings Heros to any webpage.

  • Live text selection context
  • Voice chat overlay
  • Zuno music controller
  • Syncs with main session

Backend Framework

A fast, secure, and easily self-hostable core API.

  • Django 5.2: Asynchronous core
  • Fernet: Encrypted API keys
  • Docker: 1-click self-hosting
About Us

The Heros Company

Born out of the frustration of endless API outages, Heros is built on the philosophy that your workflow should never break. We are a passionate team of engineers and AI enthusiasts dedicated to building resilient, fault-tolerant infrastructure that gracefully handles the chaos of the modern web.

Whether you're a developer building the next generation of LLM applications, or an enterprise needing guaranteed uptime, our ecosystem—from our robust Django core to our seamlessly integrated Zeno & Zuno browser extension—is designed to empower you with unbroken context, lightning-fast inference, and intelligent background capabilities. We believe in open-source collaboration, extreme reliability, and giving control back to the user.

Got Questions?

Frequently Asked Questions

Everything you need to know about the Heros AI ecosystem, failover routing, and setup.

Heros routes every query through Baymax. If Gemini (Tier 1) experiences a rate limit or API outage, Baymax instantly handshakes with OpenRouter (Tier 2), and subsequently Groq (Tier 3) if needed, preserving your chat context perfectly.
Yes, absolutely. Your API keys are encrypted at-rest using standard Fernet cryptography inside the Django database. Furthermore, keys are linked strictly to your user profile session.
Yes. Heros is released under the MIT License. You can run it locally, deploy it to staging or production using Docker, and configure custom fallback models without licensing fees.
Zeno is the core companion browser extension that puts the AI assistant directly inside any web page (right-click highlighted text lookup, voice chat overlay, etc.). Zuno is a specialized module embedded inside Zeno that allows background YouTube and music radar controls.

Bring your own keys.
Never worry about downtime again.

Free, open source, and self-hostable. Connect Gemini, OpenRouter, and Groq in under a minute.

API keys saved securely