anythink 3.1.1

A universal AI terminal workstation: multi-provider LLM chat, RAG, agentic tools, MCP, voice, 4-panel dashboard, spend tracking, batch processing, scheduling, and a plugin architecture.

Anythink

Think anything. Ask anything.

Anythink 3.1 is a universal AI terminal workstation. A Textual-powered TUI that brings RAG, agentic tools, MCP, voice, notifications, a 4-panel dashboard, a full automation and cost-awareness layer, and now a complete per-theme visual identity system to your terminal.

---

Features

Core

• Multi-provider — Groq, Google Gemini, OpenAI, Anthropic, Mistral, Cohere, Ollama, LM Studio
• Textual TUI — chat bubbles, streaming markdown, 4 color themes, persistent HUD
• Model aliases — give models friendly names (groq-fast, claude)
• Session management — auto-save, naming, undo, bookmarks, conversation branches
• File & image input — attach source files and images to any message

V2 Capabilities

• RAG — index local folders, retrieve relevant chunks, transparent 📚 Retrieved from N sources footer
• Code execution (/exec python ...) — runs code in your environment; ask/auto approval; result fed to AI
• Agentic browsing (/browse) — snippet search + full-page httpx fetch; ask/auto modes
• MCP — built-in servers (filesystem, sessions, RAG, search) + connect external MCP servers
• Dashboard mode (--dashboard or Ctrl+D) — 4 panels: session list, stats, file browser, RAG browser, tool output log
• Voice input (/voice) — mic capture → local Whisper transcription → editable text in input
• Desktop notifications — RAG build done, slow response, exec, browse, provider failure

V2.2 Visual Identity & Personalization

• Per-theme background fill — each theme (Midnight / Aurora / Ember / Arctic) paints a full-screen near-black tinted canvas, not just colored borders
• Bubble Style toggle — switch between Boxed (bordered panels) and Minimal (thin ▎ accent bar, no border, full-width transcript style) via /settings
• Role Avatars — optional ⟨Y⟩ user glyph and ✦ AI glyph next to every message header, colored in theme tones
• Compact Density mode — remove inter-bubble spacing for maximum history on screen
• Relative timestamps — just now / Xm ago / Xh ago / Yesterday / absolute fallback; live 60-second tick with /settings absolute override
• Retroactive theme application — switching theme or any visual setting instantly re-renders every existing message on screen
• Unified monochrome icon language — all emoji replaced with theme-colorable Unicode glyphs (⌕⌬▲✓✕⎘★◆) and an ASCII-safe fallback set; no more fixed-color emoji anywhere
• Collapsed session naming — the naming prompt disappears after answering, replaced by a single compact ✓ Session named: line
• Exact context-% precision — 0.3% shown for sub-1% non-zero usage; flat 0% only when context is genuinely empty

V3 Capabilities

• Per-model generation parameters — tune temperature, max tokens, top-p, frequency/presence penalty per alias (/params)
• Multi-model comparison — send one prompt to multiple aliases simultaneously; pick which response to continue with (/compare)
• Spend tracking — estimated USD cost per response, session, day, month; HUD cost indicator; optional soft budget limit (/cost)
• Prompt templates — save reusable prompts with {{variable}} placeholders; instantiate with inline args (/template, /use)
• Session export — export conversations to Markdown, JSON, or PDF (/export)
• Scheduled prompts — set recurring prompts on a cron schedule; output to file or notification (/schedule, anythink scheduler)
• Batch processing — run a file full of prompts non-interactively; parallel execution; Markdown or JSON output (anythink run)
• Self-update — check and install updates directly from PyPI (/update)
• Diagnostics — comprehensive health check: Python version, deps, API keys, local servers, config, disk (/doctor, anythink doctor)
• Config backup & restore — export your full setup as a portable JSON bundle; import on any machine (/config export|import)

---

Installation

# Base install (all providers via lazy imports, no heavy extras)
pip install anythink

Optional extras

Extra	Installs	Use case
pip install anythink[groq]	groq SDK	Groq cloud inference
pip install anythink[openai]	openai SDK	OpenAI / Azure
pip install anythink[anthropic]	anthropic SDK	Claude models
pip install anythink[gemini]	google-generativeai	Gemini Pro / Flash
pip install anythink[mistral]	mistralai SDK	Mistral models
pip install anythink[cohere]	cohere SDK	Cohere Command
pip install anythink[search]	duckduckgo-search	Free web search
pip install anythink[rag]	sentence-transformers	Local RAG embeddings
pip install anythink[voice]	openai-whisper + sounddevice	Voice input
pip install anythink[browser]	playwright	Headless web browsing
pip install anythink[mcp]	mcp SDK	External MCP server connections
pip install anythink[pdf]	fpdf2	PDF session export
pip install anythink[scheduler]	croniter	Cron-based scheduled prompts
pip install anythink[all]	Everything above	Full workstation

---

Quick Start

# Set up a provider and model alias
anythink keys add groq
anythink model add          # alias=groq-fast, provider=groq, model=llama3-8b-8192

# Start chatting (Simple Chat Mode)
anythink

# Or launch the 4-panel Dashboard
anythink --dashboard

# Run a batch of prompts non-interactively
anythink run --file prompts.txt --output results.md

# Check your installation health
anythink doctor

---

CLI Reference

# Chat
anythink                    # Simple Chat Mode
anythink --dashboard / -D   # 4-panel Dashboard Mode
anythink --version / -V     # Show version

# Key & model management
anythink keys list|add|show|update|delete|test <provider>
anythink model list|add|remove

# Plugins
anythink plugins list|info|install|remove

# V3: Batch processing (non-interactive)
anythink run --file prompts.txt --output results.md [--parallel N] [--alias A] [--format md|json]

# V3: Diagnostics
anythink doctor

# V3: Scheduled prompt automation
anythink scheduler start [--poll 60]   # Foreground loop, checks every N seconds
anythink scheduler run <name>          # Run a schedule immediately
anythink scheduler list                # Show all schedules and their status

---

Slash Commands

Chat & History

Command	Description
/help	List all commands
/clear	Clear conversation history
/history	Show recent messages
/tokens	Context window usage
/model	Active provider and model
/persona <name>	Activate a saved persona
/persona clear	Remove active persona
/exit / /quit	End the session

Sessions & Branches

Command	Description
/rename <name>	Rename current session
`/session save	load
/undo	Remove last exchange
/bookmark [turn]	Bookmark an AI response
/bookmarks	List bookmarks
/bookmark export [path]	Export bookmarks to file
/branch	Create a conversation branch
/branch list	List all branches
/branch switch <name>	Switch to branch

Files & Search

Command	Description
/file <path>	Attach a text file
/image <path>	Attach an image
/files	List pending attachments
`/search on	off

RAG

Command	Description
/rag new <name>	Create a new RAG index
/rag use <name>	Activate an index
/rag off	Deactivate RAG
/rag list	List indexes
/rag rebuild [name]	Rebuild an index
/rag info [name]	Show index details
/rag status	Show active index

Tools

Command	Description
/exec <lang> <code>	Run code in your environment
/exec mode ask|auto	Change exec approval mode
/browse <url|query>	Fetch web page or search
/browse mode ask|auto|http|headless	Change browse mode

MCP

Command	Description
/mcp list	List registered servers
/mcp tools	List available tools
/mcp call <tool> [k=v ...]	Call a tool
/mcp connect <name> stdio <cmd>	Connect to external server
/mcp disconnect <name>	Disconnect
/mcp server start|stop|status	Run Anythink as MCP server

Voice & Notifications

Command	Description
/voice	Start recording (Enter to stop)
/voice model tiny|base|small|medium|large|turbo	Set Whisper model
/voice language <code>	Pin transcription language
/notify on|off|status	Notification control
/notify type <name> on|off	Per-type toggle

V3 — Model Parameters & Spend

Command	Description
/params	Show generation params for the active alias
/params temperature=0.3 max_tokens=2048 top_p=0.9	Set params (key=value pairs)
/params reset	Reset to provider defaults
/cost	Session spend estimate
/cost today	Today's spend across all sessions
/cost month	This month's spend
/cost by-model	Lifetime spend broken down per model
/cost by-provider	Lifetime spend broken down per provider

V3 — Templates

Command	Description
/template save <name> <body>	Save a prompt template (supports {{variable}})
/template list	List all saved templates
/template show <name>	Preview a template
/template delete <name>	Remove a template
/use <name> [key=value ...]	Render a template and send as next message

V3 — Export & Compare

Command	Description
/export	Export session as Markdown (default)
/export json [path]	Export as JSON
/export pdf [path]	Export as PDF (pip install anythink[pdf])
/export markdown --range 1-10	Export a turn range
/compare <alias1> <alias2> [...]	Compare up to 4 models on the next prompt

V3 — Scheduling

Command	Description
/schedule list	List all scheduled prompts
/schedule add <name> "<cron>" <prompt>	Create a scheduled prompt
/schedule enable <name>	Enable a paused schedule
/schedule disable <name>	Pause a schedule
/schedule remove <name>	Delete a schedule
/schedule run <name>	Run a schedule immediately

V2.2 — Visual Settings

Command	Description
/settings	Open the interactive settings overlay
/theme <name>	Switch theme instantly (midnight | aurora | ember | arctic)

All five new visual options (Bubble Style, Role Avatars, Density, Timestamps, Icon Style) are adjustable live from the /settings overlay — every change takes effect immediately and retroactively across the full conversation.

V3 — Maintenance

Command	Description
/doctor	Full installation health check
/update check	Check for a newer version on PyPI
/update	Upgrade Anythink to the latest version
/config export [path]	Export config as a portable JSON bundle
/config import <path>	Restore config from a bundle

Dashboard Shortcuts

Key	Action
Ctrl+D	Toggle Simple / Dashboard mode
Ctrl+L	Toggle Sessions panel
Ctrl+R	Toggle Stats panel
Tab	Cycle bottom tabs
Escape	Return focus to input

---

Configuration

All data stored under the XDG Base Directory hierarchy:

Path	Default
Config	~/.config/anythink/config.yaml
Models	~/.config/anythink/models.yaml
Templates	~/.config/anythink/templates.yaml
Schedules	~/.config/anythink/schedules.yaml
Sessions	~/.local/share/anythink/sessions/
Exports	~/.local/share/anythink/exports/
Spend log	~/.local/share/anythink/spend.yaml
RAG metadata	~/.local/share/anythink/rag/
RAG vectors	~/.cache/anythink/rag/

Key config.yaml options:

active_theme: midnight          # midnight | aurora | ember | arctic
default_model_alias: groq-fast
session_autosave: true
search_provider: duckduckgo
embedding_backend: local        # local | api
exec_mode: ask                  # ask | auto
browse_mode: http               # http | headless
browse_autonomy: ask            # ask | auto
voice_model: base               # tiny|base|small|medium|large|turbo

# V2.2 visual identity (all adjustable live via /settings)
bubble_style: boxed             # boxed | minimal
density: comfortable            # comfortable | compact
show_avatars: false             # true | false
timestamps: relative            # relative | absolute
icon_style: unicode             # unicode | ascii

# V3 spend tracking
spend_tracking: true
spend_budget_period: monthly    # monthly | daily
spend_budget_soft_limit: null   # e.g. 10.0 for a $10 soft limit

---

Development

git clone https://github.com/dhineshtheprogrammer/anythink_ai
cd anythink_ai
pip install -e ".[all,dev]"

ruff check src/
black --check src/ tests/
mypy src/anythink
bandit -r src/anythink -c pyproject.toml
PYTHON_KEYRING_BACKEND=keyrings.alt.file.PlaintextKeyring pytest tests/ -v
python -m build

See CONTRIBUTING.md for contribution guidelines and CHANGELOG.md for release history.

---

License

MIT — see LICENSE for details.