Hugging Face's logo

Spaces:
muralipala
/
deepshell
Running

App Files Community
deepshell / README.md
muralipala1504
Fix: specify Dockerfile.hf in README for HF build
3d52446
preview code
|
Raw
History Blame Contribute Delete
11.1 kB
metadata 
title: DeepShell
emoji: 💻
colorFrom: green
colorTo: blue
sdk: docker
dockerfile: Dockerfile.hf
app_port: 7860
license: mit
short_description: AI DevOps assistant with Hindi TTS — Linux, Docker, K8s
pinned: false
tags:
  - devops
  - linux
  - docker
  - kubernetes
  - fastapi
  - tts
  - hindi
  - sysadmin
  - llm
  - groq
  - piper
  - infrastructure
  - cloud
  - ansible
  - terraform

💻 DeepShell ModUI

Python FastAPI Docker License LLM TTS

AI-powered DevOps assistant AND trainer — specialized for Linux SysAdmin, DevOps, Docker, Kubernetes, IaC, and Cloud operations. Not a generic ChatGPT wrapper. Strictly focused. Blazing fast. Runs 100% offline with Ollama.

Real-time streaming • 3 AI modes • Multilingual Voice TTS • Syntax highlighting • Light/Dark theme • Docker ready • Groq + Cerebras + Ollama

🚀 What is DeepShell ModUI?

DeepShell ModUI is a fast, minimal web application that brings AI assistance directly into your infrastructure workflows:

  • Frontend: Responsive chat UI — Vanilla JS, zero frameworks, light/dark theme
  • Backend: FastAPI with SSE streaming + /translate + /tts endpoints
  • LLM: Groq (default) → Cerebras (failover) → Ollama (offline)
  • Smart Context: Maintains conversation history for better responses
  • Strict Scope: Refuses off-topic questions — stays on DevOps/SysAdmin only
  • Three Modes: Assistant, Trainer, and Dry Run
  • Multilingual TTS: Piper (offline, self-hosted) for Hindi + Web Speech API for English

Perfect for DevOps engineers, SysAdmins, Cloud admins, and IaC Specialists who need accurate commands and scripts — fast.

🤖 Three Modes — One Tool

🔵 Assistant Mode (Default)

Fast, direct answers. Commands, scripts, and explanations for DevOps tasks.

🎓 Trainer Mode

Structured teaching format — ideal for learning and upskilling. Every response follows a consistent teaching structure with What it does, Command/Example, Breaking it down, Common mistakes, Try this next, and Pro tip sections.

🔍 Dry Run Mode

Safety-first command analysis — see exactly what WOULD happen before running anything. Shows risk level and safe alternatives.

Toggle between modes instantly using the Assistant / Trainer / Dry Run buttons in the header.

✨ Features

Free Version

  • ✅ Assistant Mode — fast, direct DevOps answers
  • ✅ Trainer Mode — structured teaching with examples and pro tips
  • ✅ Dry Run Mode — safety-first command analysis before execution
  • ✅ 🔊 Multilingual Voice TTS — English (Web Speech API) + Hindi via Piper (self-hosted, offline)
  • ✅ 🌐 Language selector — EN / HI / TA / TE / AR in header dropdown
  • ✅ 🔁 Replay — replay last response anytime
  • ✅ 🌙/☀️ Theme toggle — light and dark mode with persistence
  • ✅ Real-time streaming responses via SSE
  • ✅ Syntax highlighting — Bash, Python, Dockerfile, YAML, HCL (Terraform)
  • ✅ Session context — AI remembers your conversation
  • ✅ Copy buttons on all code blocks
  • ✅ Export chat as Markdown file
  • ✅ Clear chat to reset session
  • ✅ Keyboard shortcuts (Ctrl+Enter, Ctrl+K, Escape)
  • ✅ Green status indicator — live backend health check
  • ✅ Strict AI scope — politely refuses off-topic questions
  • ✅ Works locally or over LAN
  • ✅ Docker Compose — DeepShell + LibreTranslate as sidecar
  • ✅ Start/Stop scripts — no terminal lock
  • ✅ Groq — blazing fast cloud inference
  • ✅ Cerebras — failover provider
  • ✅ Ollama — run 100% offline, no API key needed

Coming Soon — Pro Version 💎

  • 🔜 One-click command execution with approval flow
  • 🔜 Persistent chat history (database)
  • 🔜 Multi-host support (remote server connections)
  • 🔜 Workflow automation and command chaining
  • 🔜 File upload — analyze logs and configs
  • 🔜 Chatterbox TTS — cloned voice as your personal instructor
  • 🔜 Quiz mode for certification prep (RHCSA, CKA, AWS)
  • 🔜 Stop audio on language switch

🛠️ Tech Stack

Layer Technology
Frontend HTML5, CSS3, Vanilla JavaScript
Backend FastAPI, Uvicorn, SSE
LLM (Cloud) Groq API — llama-3.3-70b-versatile
LLM (Failover) Cerebras — llama3.1-8b
LLM (Local) Ollama — phi3, qwen2.5-coder, llama3 etc
Syntax Prism.js (dark + light themes)
Markdown marked.js
TTS (English) Web Speech API (browser native)
TTS (Hindi) Piper TTS — hi_IN-rohan-medium (self-hosted, offline)
Translation LibreTranslate (self-hosted, en→hi)
Python 3.11+

🌐 Multilingual TTS Architecture

DeepShell uses a self-hosted translation + TTS pipeline — no cloud APIs, no rate limits, no data leaving your network:

  • English: Web Speech API (browser built-in, zero latency)
  • Hindi: LibreTranslate → Piper TTS (offline, self-hosted)
  • Sentence-by-sentence: Audio starts immediately, no full-response wait
  • Fallback: If backend TTS fails, falls back to Web Speech API automatically

⚠️ Note: Switching language mid-playback does not stop current audio. This will be fixed in a future release.

🖼️ Preview

🌐 Live Demo: deepshell.cloud — try it directly in your browser!

⚡ Quick Start (Python — No Docker)

Prerequisites

  • Python 3.11+
  • Git
  • Groq API key — Get one free OR Ollama for offline use

1. Clone and Install 

git clone https://github.com/muralipala1504/deepshell_modui.git
cd deepshell_modui

pip install -r deepshell-backend/requirements.txt
pip install httpx
pip install -e deepshell-backend/

2. Set Groq API Key 

echo 'export GROQ_API_KEY="your_groq_api_key_here"' >> ~/.bashrc
source ~/.bashrc

3. Install Piper TTS (Hindi voice) 

cd ~
wget https://github.com/rhasspy/piper/releases/download/2023.11.14-2/piper_linux_x86_64.tar.gz
tar -xzf piper_linux_x86_64.tar.gz

mkdir -p ~/piper/voices
wget https://huggingface.co/rhasspy/piper-voices/resolve/main/hi/hi_IN/rohan/medium/hi_IN-rohan-medium.onnx -O ~/piper/voices/hi_IN-rohan-medium.onnx
wget https://huggingface.co/rhasspy/piper-voices/resolve/main/hi/hi_IN/rohan/medium/hi_IN-rohan-medium.onnx.json -O ~/piper/voices/hi_IN-rohan-medium.onnx.json

4. Install LibreTranslate (Hindi translation) 

python3 -m venv ~/.venvs/libretranslate
source ~/.venvs/libretranslate/bin/activate
pip install libretranslate==1.9.5
deactivate

5. Firewall (remote/LAN access only) 

# firewalld (RHEL/CentOS/Fedora/AlmaLinux)
sudo firewall-cmd --add-port=8001/tcp --permanent
sudo firewall-cmd --add-port=5000/tcp --permanent
sudo firewall-cmd --reload

# ufw (Ubuntu/Debian)
sudo ufw allow 8001/tcp
sudo ufw allow 5000/tcp

6. Start / Stop 

# Start DeepShell + LibreTranslate in background
./start.sh

# Check logs
tail -f deepshell.log
tail -f libretranslate.log

# Stop everything
./stop.sh

7. Open in Browser

🎯 What DeepShell Excels At

Domain Examples
🐳 Docker Dockerfiles, compose files, container management
☸️ Kubernetes kubectl commands, YAML manifests, troubleshooting
🔧 Linux SysAdmin System commands, log analysis, performance tuning
🏗️ IaC Terraform (HCL), Ansible, CloudFormation
☁️ Cloud AWS CLI, best practices
📜 Scripting Bash, Python automation for ops tasks

DeepShell politely refuses general knowledge, entertainment, or anything outside DevOps/SysAdmin scope.

🔒 Security & Safety

  • ✅ No automatic execution — commands displayed with copy buttons only
  • ✅ Dry Run mode — analyze before you execute
  • ✅ User control — you decide what runs on your system
  • ✅ Session isolation — each browser session is independent
  • ✅ No data persistence — conversations in-memory only (free version)
  • ✅ Rate limiting — built-in via slowapi
  • ✅ Strict AI scope — refuses off-topic requests
  • ✅ Ollama option — LLM data never leaves your network
  • ✅ Self-hosted TTS — Piper runs locally, no audio sent to cloud

⌨️ Keyboard Shortcuts

Shortcut Action
Ctrl + Enter Send message
Ctrl + K Clear chat
Escape Focus input

🛣️ Roadmap

Phase 1 — Free Version ✅ Complete

  • ✅ SSE streaming + session context
  • ✅ Syntax highlighting (Bash, Python, Dockerfile, YAML, HCL)
  • ✅ Copy buttons + Export chat as Markdown
  • ✅ Docker Compose support + Start/Stop scripts
  • ✅ Assistant / Trainer / Dry Run modes
  • ✅ Multilingual Voice TTS — English + Hindi (Piper, offline)
  • ✅ Self-hosted translation via LibreTranslate
  • ✅ Backend /translate + /tts endpoints
  • ✅ Replay + Light/Dark theme
  • ✅ Groq + Cerebras + Ollama provider support

Phase 2 — Pro Version 💎 Next

  • 🔜 One-click command execution with approval flow
  • 🔜 Persistent history (database)
  • 🔜 Chatterbox TTS — cloned voice instructor
  • 🔜 Quiz mode for certification prep (RHCSA, CKA, AWS)
  • 🔜 Multi-session management
  • 🔜 Workflow automation
  • 🔜 Stop audio on language switch

Phase 3 — Advanced

  • 🔜 File upload (analyze logs, configs)
  • 🔜 Multi-host support
  • 🔜 Kubernetes dashboard integration
  • 🔜 Additional Indian languages (Tamil, Telugu)

🤝 Contributing

Contributions welcome! Please:

  • Keep code simple and maintainable
  • Focus on DevOps/SysAdmin/IaC use cases
  • Follow existing code style
  • Test before submitting PRs

📝 License

MIT License — see LICENSE for details.

🙏 Acknowledgments

📧 Contact

Maintainer: @muralipala1504 Issues: GitHub Issues Website: deepshell.cloud ✅ LIVE

Built with ❤️ for DevOps engineers who value speed, accuracy, and control.