ANARCHY v1.1.1 — Maintenance & Cleanup

ANARCHY v1.1.1 — Maintenance & Cleanup

No cloud. No telemetry. No gods. Only Zul.

What changed

Removed 15 stray/empty files left behind by audit scripts:
'content', 'tool_name', 0.82, ANARCHY (root file), app.hardware, app.memory,
console, core.text, from, import_test.txt, individual,
install_check_out.txt, install_test_out.txt, intent_confidence, project_root, total_files

Fixed IndentationError in test/test_safety.py — test_self_modify_is_high_risk had a stale comment block that created a nested def with no body. CI is now green across Python 3.10/3.11/3.12 on Linux and Windows.

Updated pyproject.toml: version bumped to 1.1.1, author metadata updated to include GitHub username and email.

Updated pytest dependency to >=9.0.3 via Dependabot.

---

Install

# Linux / macOS
curl -fsSL https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.ps1 | iex

# pip (coming soon via PyPI)
pip install anarchy-ai

ANARCHY v1.1.0 — Domain Consolidation & 3-Phase Installer

ANARCHY v1.1.0

NO CLOUD. NO TELEMETRY. NO GODS. ONLY ZUL.

---

Install

# Windows
irm https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.ps1 | iex

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.sh | bash

# Android (Termux)
pkg install git python cmake clang curl
curl -fsSL https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.sh | bash

---

What Changed Since v1.0.0

✨ 3-Phase Hard-Gated Installer

The installer now has three strict phases that must each pass 100% before the next begins:

• Phase 1 DETECT — stdlib only, cannot fail. Reads OS, CPU/RAM/GPU, CUDA, ROCm, Metal. Selects optimal model and context size automatically.
• Phase 2 INSTALL — Tracks every step. Any critical failure prints an itemised report and stops before Phase 3.
• Phase 3 BUILD — Compiles all Python files and verifies core imports before suggesting launch.

Hardware-adaptive model selection across 7 VRAM tiers: from llama3.2:1b (CPU-only) up to llama3:70b (25+ GB VRAM).

✨ Cloud Provider Hybrid Routing

Optional cloud acceleration — everything stays local by default:

• Groq (300+ tok/s), OpenRouter (DeepSeek-R1 + 50 free models), Gemini (1M tok/day free), Cerebras (1000+ tok/s)
• provider add <name> stores keys in encrypted vault, never in plain text
• Auto mode: uses cloud when available, silently falls back to local

✨ New Subsystems (13 added since v1.0.0)

ContextCompactor, IntentRouter, StructuredOutput, SubAgentRunner, AgentCoordinator, TokenTracker, MemDir, MemoryRouterBridge, ProviderRegistry, Migrator, ProjectState, AutoDoc, DebugCLI

🔧 Domain Module Consolidation

34 single-purpose files collapsed into 9 domain modules. core/main.py reduced from 3,110 → 899 lines via a core/handlers/ package (10 focused modules).

New module	Absorbed
core/llm.py	llm_manager + model_router + token_tracker + cache_manager + connectivity + context_compactor + structured_output
core/cognition.py	nlp_engine + reasoner + dialogue_state_tracker + intent_router
core/execution.py	executor + task_scheduler + subagent + multi_agent
core/extensions.py	plugin_system + skill_installer + persona_engine
core/project.py	project_indexer + project_state + discovery_engine + knowledge_base + blueprint
core/monitoring.py	quality_metrics + performance_manager + health + watchdog
core/dev_tools.py	git_wizard + hook_manager + autodoc + lsp_client
core/text.py	formatter + system_prompt
core/Config.py	+ secrets + migrations

🐛 Bug Fixes

• lsp_client.py: SERVERS dict closed prematurely — go, c, ruby, lua were orphaned outside the dict literal
• nlp_engine.py: _extract_temporal_expressions() called but never defined; implemented with full regex extractor
• UTF-8 BOM: Stripped from Windows-authored files causing SyntaxError on Linux/macOS
• Intent routing: Knowledge questions no longer incorrectly routed to tool execution
• Windows 11: Correctly identified (build ≥ 22000), not misreported as Windows 10
• Termux: faiss-cpu and pyautogui auto-skipped; llama-cpp built with LLAMA_NATIVE=OFF

---

Full changelog: CHANGELOG.md

ANARCHY v1.0.0

ANARCHY v1.0.0 — Initial Release

NO CLOUD. NO TELEMETRY. NO GODS. ONLY ZUL.

ANARCHY is a fully local, offline-first AI shell. It runs entirely on your hardware via Ollama and GGUF models — no API keys, no subscriptions, no data ever leaving your machine.

---

Install in One Command

Windows

irm https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.ps1 | iex

macOS / Linux / Chromebook

curl -fsSL https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.sh | bash

Android (Termux)

pkg install git python curl && curl -fsSL https://raw.githubusercontent.com/psfr4590-afk/ANARCHY/main/get-anarchy.sh | bash

Or from this release, download the wheel and run:

pip install anarchy-1.0.0-py3-none-any.whl
anarchy

---

What's in v1.0.0

Core Engine (40 subsystems)

• Local LLM — Ollama primary (llama3.2:3b default) with llama-cpp-python GGUF fallback
• NLP/NLU Pipeline — intent classification, entity extraction, sentiment, urgency scoring
• Tool Execution — 20+ tools behind a 4-tier permission gate (SAFE / CONFIRM / HIGH_RISK / DENIED)
• Semantic Memory — vector memory (FAISS) + JSON episodic memory across sessions
• Smart Model Routing — complexity scoring 0–4, auto-selects the right model size per query
• RAG Search — index any codebase with sentence-transformers + FAISS, query in plain English
• Task Scheduler — cron-style daemon, schedule LLM prompts to run daily/hourly/weekly
• Voice Input — local Whisper transcription, no cloud, works offline
• Skill Installer — install community skills from any GitHub URL, hot-reload, no restart needed

New in This Release

• LSP Code Intelligence (diagnose <file>) — JSON-RPC Language Server Protocol client. Runs real language servers locally (pyright for Python, clangd for C/C++, rust-analyzer, gopls for Go, typescript-language-server). Returns actual static analysis errors, explained by your local LLM.
• Smart Git Workflow (commit, commit-push) — stages your diff, feeds it to Ollama with the last 10 commit messages for style matching, and auto-drafts the commit message. Blocks on sensitive file patterns (.env, credentials.json). Full commit → push → PR in one command.
• Git Hook Manager (hooks list/install-anarchy/remove/show) — install, remove, and inspect git hooks. hooks install-anarchy adds a pre-commit hook that runs LSP diagnostics on staged files before every commit and gives you a chance to abort.
• Built-in Review Skills — security_review (7-category security audit: injection, auth, secrets, crypto, deps, error handling, input validation) and code_review (6-dimension review: comments, test gaps, silent failures, types, maintainability, simplification). Invoked via natural language.

Cross-Platform Support

• Windows 10/11 (x86_64, GTX/RTX GPU or CPU)
• macOS Intel (x86_64)
• macOS Apple Silicon (M1/M2/M3 — arm64, Homebrew at /opt/homebrew)
• Linux (Ubuntu/Debian apt, Fedora dnf, Arch pacman)
• Chromebook via Crostini (Linux container)
• Android via Termux (no Ollama binary; GGUF fallback or remote Ollama)

Test Suite

69 tests, all passing on Python 3.10 / 3.11 / 3.12.

---

What's Coming

• PyPI publication (pip install anarchy)
• Plugin system for community-built tool packs
• Web UI mode (serve the REPL over localhost)
• Multi-agent workflows (one ANARCHY orchestrates others)

---

Requirements

• Python 3.10+
• Ollama (auto-installed by the one-liner scripts)
• 8 GB RAM minimum (4 GB works with phi3:mini)
• GPU optional — runs fine on CPU

---

MIT License · No cloud · No telemetry · No gods