dolios-agent

<p align="center">
  <img src="assets/banner.png" alt="Dolios Agent" width="100%">
</p>

# Dolios Agent Δ

<p align="center">
  <a href="https://github.com/guglxni/dolios-agent/blob/main/SECURITY-AUDIT.md"><img src="https://img.shields.io/badge/OWASP-2025%20Audited-FFD700?style=for-the-badge" alt="OWASP Audited"></a>
  <a href="https://github.com/guglxni/dolios-agent"><img src="https://img.shields.io/badge/Tests-56%20Passing-brightgreen?style=for-the-badge" alt="Tests"></a>
  <a href="https://github.com/guglxni/dolios-agent/blob/main/LICENSE"><img src="https://img.shields.io/badge/License-MIT-green?style=for-the-badge" alt="License: MIT"></a>
  <a href="https://github.com/guglxni"><img src="https://img.shields.io/badge/Built%20by-Aaryan%20Guglani-blueviolet?style=for-the-badge" alt="Built by Aaryan Guglani"></a>
</p>

**The Crafty Agent: Scheme. Execute. Deliver.** Dolios is a self-improving AI agent with production-grade sandboxing. It builds on [Hermes Agent](https://github.com/NousResearch/hermes-agent)'s closed learning loop and [NemoClaw](https://github.com/NVIDIA/NemoClaw)'s Landlock/seccomp isolation, adding multi-provider inference routing, [AI-DLC](https://github.com/awslabs/aidlc-workflows) methodology awareness, and a [DSPy/GEPA](https://github.com/NousResearch/hermes-agent-self-evolution) self-evolution pipeline. Runs on a $5 VPS, a GPU cluster, or serverless.

Use any model: [NVIDIA Nemotron](https://build.nvidia.com), [Nous Portal](https://portal.nousresearch.com), [OpenRouter](https://openrouter.ai) (200+ models), [Kimi/Moonshot](https://platform.moonshot.ai), [MiniMax](https://www.minimax.io), OpenAI, Anthropic, or local via Ollama/vLLM. Switch with `dolios model`, no code changes, no lock-in.

<table>
<tr><td><b>Sandboxed execution</b></td><td>Every tool call runs inside a NemoClaw/OpenShell sandbox with Landlock filesystem isolation, seccomp process restrictions, and deny-by-default network policies. No tool can <code>rm -rf /</code> or exfiltrate data; the sandbox blocks it at the kernel level.</td></tr>
<tr><td><b>Self-improving skills</b></td><td>Closed learning loop: the agent creates procedural skills from experience, improves them via DSPy/GEPA optimization ($2-10 per cycle, no GPU), and persists knowledge across sessions. Compatible with the <a href="https://agentskills.io">agentskills.io</a> open standard.</td></tr>
<tr><td><b>Lives where you do</b></td><td>Telegram, Discord, Slack, WhatsApp, Signal, and CLI, all from a single gateway process. Voice memo transcription, cross-platform conversation continuity.</td></tr>
<tr><td><b>Methodology-aware</b></td><td>AI-DLC workflow engine: Inception → Construction → Operations phases for structured task execution. The agent asks the right questions before coding, follows a plan, and validates before delivering.</td></tr>
<tr><td><b>Multi-provider inference</b></td><td>Smart routing across NVIDIA, Nous, OpenRouter, OpenAI, Anthropic, Kimi, MiniMax, and local models, all intercepted by the sandbox inference gateway. Cost-aware selection based on task type.</td></tr>
<tr><td><b>Security-first</b></td><td>OWASP Top 10:2025 + LLM Top 10:2025 audited. Fail-closed defaults, prompt injection scanning, API key redaction, atomic writes, 7 evolution constraint gates. 67/67 findings resolved.</td></tr>
<tr><td><b>Research-ready</b></td><td>Batch trajectory generation, Atropos RL environments, execution trace collection, trajectory compression for training the next generation of tool-calling models.</td></tr>
</table>

---

## Quick Install

```bash
curl -fsSL https://raw.githubusercontent.com/guglxni/dolios-agent/main/scripts/install.sh | bash
```

Works on Linux, macOS, and WSL2. The installer handles Python, uv, dependencies, vendor repos, and the `dolios` command.

> **Windows:** Native Windows is not supported. Please install [WSL2](https://learn.microsoft.com/en-us/windows/wsl/install) and run the command above.

After installation:

```bash
source ~/.bashrc    # reload shell (or: source ~/.zshrc)
dolios setup        # guided configuration wizard
dolios              # start chatting!
```

Or install manually:

```bash
git clone https://github.com/guglxni/dolios-agent.git
cd dolios-agent
bash scripts/install.sh    # clones vendor repos + installs deps
dolios setup
dolios
```

---

## Getting Started

```bash
dolios              # Interactive CLI — start a conversation
dolios setup        # Full setup wizard (providers, API keys, sandbox)
dolios model        # Choose your inference provider and model
dolios sandbox status   # Check sandbox health and policy state
dolios sandbox policy   # View active network policies
dolios sandbox approve  # Approve pending endpoint access requests
dolios evolve run   # Run self-evolution pipeline on a skill
dolios aidlc        # Show current AI-DLC workflow phase
dolios doctor       # Diagnose any issues
```

---

## Architecture

```
┌──────────────────────────────────────────────────────────────┐
│                      USER INTERFACES                         │
│    CLI (TUI) · Telegram · Discord · Slack · WhatsApp         │
└──────────┬───────────────────────────────────────────────────┘
           │
           ▼
┌──────────────────────────────────────────────────────────────┐
│                    DOLIOS ORCHESTRATOR                       │
│                                                              │
│  Policy Bridge · Inference Router · AI-DLC Engine            │
│  Brand Layer   · Trace Collector  · Prompt Injection Scan    │
└──────────┬───────────────────────────────────────────────────┘
           │
           ▼
┌──────────────────────────────────────────────────────────────┐
│                   HERMES AGENT RUNTIME                       │
│                                                              │
│  Agent Loop · Skills · Memory · Tools · Gateway              │
│  Subagents  · Honcho · Cron   · MCP   · FTS5 Search          │
└──────────┬───────────────────────────────────────────────────┘
           │
           ▼
┌──────────────────────────────────────────────────────────────┐
│                   NEMOCLAW SANDBOX LAYER                     │
│                                                              │
│  Network Policy · Filesystem Isolation · Process Sandbox     │
│  (deny-default) · (Landlock strict)    · (seccomp)           │
│  Inference Gateway · Blueprint Lifecycle · Policy YAML       │
└──────────┬───────────────────────────────────────────────────┘
           │
           ▼
┌──────────────────────────────────────────────────────────────┐
│                    INFERENCE PROVIDERS                       │
│                                                              │
│  NVIDIA Nemotron · Nous Portal · OpenRouter (200+ models)    │
│  Kimi / Moonshot · MiniMax     · OpenAI · Local (Ollama)     │
└──────────────────────────────────────────────────────────────┘
```

---

## Documentation

| Section | What's Covered |
|---------|---------------|
| [Quickstart](#quick-install) | Install → setup → first conversation in 2 minutes |
| [Technical PRD](https://github.com/guglxni/dolios-agent/blob/main/dolios-technical-prd.md) | Full architecture spec, integration maps, implementation roadmap |
| [AI-DLC Workflow](https://github.com/guglxni/dolios-agent/blob/main/CLAUDE.md) | Workflow rules, code conventions, security rules (DOLIOS-SEC) |
| [Agent Instructions](https://github.com/guglxni/dolios-agent/blob/main/AGENTS.md) | Instructions for AI coding agents working on this repo |
| [Security Audit](https://github.com/guglxni/dolios-agent/blob/main/SECURITY-AUDIT.md) | OWASP 2025 + LLM Top 10 audit, 67 findings, all resolved |
| [Contributing](https://github.com/guglxni/dolios-agent/blob/main/CONTRIBUTING.md) | Development setup, PR process, skill/policy authoring |
| [Brand Identity](https://github.com/guglxni/dolios-agent/tree/main/brand) | SOUL.md personality, voice guidelines, brand context |
| [Skills](https://github.com/guglxni/dolios

[truncated…]