@robbiesrobotics/alice-agents 1.5.9 → 1.5.11

package/README.md

@@ -1,187 +1,207 @@
-# 🧠 A.L.I.C.E. — AI Agent Framework for NemoClaw & OpenClaw
+# 🧠 A.L.I.C.E. — Multi-Agent AI Team Framework
 
 **Adaptive Learning & Intelligent Coordination Engine**
 
-One conversation. One orchestrator. Ten starter agents — with 23 more in Pro. A.L.I.C.E. turns your agent runtime into a full AI team — talk to A.L.I.C.E. (or just Alice), and she routes your request to the right expert.
+One conversation. One orchestrator. 10 starter agents — with 23 more in Pro. Tell A.L.I.C.E. what you need, and she routes it to the right specialist and delivers the result.
 
-**NemoClaw compatible** — A.L.I.C.E. v1.2.7+ runs natively on [NVIDIA NemoClaw](https://nvidia.com/nemoclaw), the secure open-source agent runtime. Agents execute inside the OpenShell sandbox for enterprise-grade security.
+A.L.I.C.E. runs on any supported runtime — OpenClaw, NemoClaw, or Hermes Agent. No required API key, no vendor lock-in.
+
+## Supported Runtimes
+
+| Runtime | Status | Install Guide |
+|---------|--------|---------------|
+| **Hermes Agent** | ✅ Recommended | [hermes-agent.nousresearch.com](https://hermes-agent.nousresearch.com) |
+| **NemoClaw** | ✅ Supported | [nvidia.com/nemoclaw](https://nvidia.com/nemoclaw) |
+| **OpenClaw** | ✅ Supported | [openclaw.ai](https://openclaw.ai) |
+
+## Prerequisites
+
+**Before installing**, make sure your runtime has a default model set:
+
+- **Hermes Agent:** Run `hermes model` to select a provider and model
+- **NemoClaw / OpenClaw:** Run `openclaw configure` → select "model" to configure your default
+
+A.L.I.C.E. uses whatever model your runtime already has configured. If no model is set, the installer will tell you exactly which command to run.
 
 ## Quick Start
 
 ```bash
+# Interactive install — auto-detects your runtime
 npx @robbiesrobotics/alice-agents
-```
 
-That's it. The installer detects your runtime (NemoClaw or OpenClaw) and sets everything up. If neither is installed, it will offer to install NemoClaw — the recommended option.
+# Non-interactive (Starter tier, auto-detected defaults)
+npx @robbiesrobotics/alice-agents --yes
+
+# Force a specific runtime
+npx @robbiesrobotics/alice-agents --yes --runtime hermes
+npx @robbiesrobotics/alice-agents --yes --runtime openclaw
+```
 
 ## What You Get
 
-**Starter** includes 10 agents. **Pro** unlocks 23 more — [sign up at getalice.av3.ai](https://getalice.av3.ai/signup?plan=pro)
-
-**A.L.I.C.E. | Control Cloud** is available as a Pro add-on. If enabled during install, the package now:
-- installs the `mission-control-bridge` plugin into your OpenClaw home
-- writes a portable local A.L.I.C.E. | Control config at `~/.openclaw/.alice-mission-control.json`
-- enables the bridge in `openclaw.json` so your runtime can forward live telemetry to A.L.I.C.E. | Control
-- installs a bundled `coding-agent` skill that prefers Codex for OpenAI defaults and Claude Code for Anthropic defaults
-
-An orchestrator (A.L.I.C.E., also addressable as Alice or Olivia) backed by specialist agents across every domain:
-
-| Agent | Domain | Emoji | Tier |
-|-------|--------|-------|------|
-| **A.L.I.C.E.** | Orchestration | 🧠 | Starter |
-| **Dylan** | Development | 💻 | Starter |
-| **Selena** | Security | 🛡️ | Starter |
-| **Devon** | DevOps | 🚀 | Starter |
-| **Quinn** | QA/Testing | ✅ | Starter |
-| **Felix** | Frontend | 🖥️ | Starter |
-| **Daphne** | Documentation | 📝 | Starter |
-| **Rowan** | Research | 🔍 | Starter |
-| **Darius** | Data | 📊 | Starter |
-| **Sophie** | Support | 💬 | Starter |
-| **Hannah** | HR | 👥 | Pro |
-| **Aiden** | Analytics | 📈 | Pro |
-| **Clara** | Communication | ✍️ | Pro |
-| **Avery** | Automation | ⚙️ | Pro |
-| **Owen** | Operations | 🔧 | Pro |
-| **Isaac** | Integration | 🔌 | Pro |
-| **Tommy** | Travel | ✈️ | Pro |
-| **Sloane** | Sales | 💼 | Pro |
-| **Nadia** | UI/UX Design | 🎨 | Pro |
-| **Morgan** | Marketing | 📣 | Pro |
-| **Alex** | API Crawling | 🕷️ | Pro |
-| **Uma** | UX Research | 🧪 | Pro |
-| **Caleb** | CRM | 🗂️ | Pro |
-| **Elena** | Estimation | 📋 | Pro |
-| **Audrey** | Accounting | 💰 | Pro |
-| **Logan** | Legal | ⚖️ | Pro |
-| **Eva** | Executive Assistant | 📌 | Pro |
-| **Parker** | Project Management | 📅 | Pro |
-| **Aria** | Autonomous Research | 🔬 | Pro |
-| **Nate** | n8n Automation | 🔧 | Pro |
-| **MaxxiPro** | Roof Maxx Expert | 🏠 | Pro |
-| **AccuScope** | AccuLynx CRM Auditor | 📊 | Pro |
-| **SmokeTestAgent** | QA Specialist | 🧪 | Pro |
-
-## Model Flexibility
-
-A.L.I.C.E. works with **whatever model you already have configured** in OpenClaw/NemoClaw. There's no required API key — just use what you've got.
-
-When you install, the installer will **auto-detect your configured model** and let the agent team inherit that runtime default. You can also choose from presets or specify a custom model for the OpenClaw/NemoClaw defaults A.L.I.C.E. should use.
-
-| Preset | Models | Best For |
-|--------|--------|----------|
-| **Auto-detect** (default) | Your current OpenClaw/NemoClaw default | Zero friction — agents inherit what your runtime already uses |
-| **Sonnet** | claude-sonnet-4-6 for all | Balanced speed + quality (requires Anthropic key) |
-| **Opus + Sonnet** | Opus for orchestrator, Sonnet for specialists | Maximum quality (requires Anthropic key) |
-| **OpenAI** | GPT-5.4 / GPT-5.4-mini | OpenAI users |
-| **Local (Ollama)** | Local models | Privacy, offline use, no API key needed |
-| **Custom** | Your choice | Full control |
-
-> **Tip:** We prefer Claude Opus for orchestration when available — but A.L.I.C.E. is model-agnostic. Whatever model OpenClaw has configured will work.
+**Starter** — 10 core agents. **Pro** — 23 additional agents. [Sign up at getalice.av3.ai](https://getalice.av3.ai/signup?plan=pro)
+
+| Agent | Domain | Tier |
+|-------|--------|------|
+| **A.L.I.C.E.** | Orchestration | Starter |
+| **Dylan** | Development | Starter |
+| **Selena** | Security | Starter |
+| **Devon** | DevOps | Starter |
+| **Quinn** | QA / Testing | Starter |
+| **Felix** | Frontend | Starter |
+| **Daphne** | Documentation | Starter |
+| **Rowan** | Research | Starter |
+| **Darius** | Data | Starter |
+| **Sophie** | Support | Starter |
+| **Hannah** | HR | Pro |
+| **Aiden** | Analytics | Pro |
+| **Clara** | Communication | Pro |
+| **Avery** | Automation | Pro |
+| **Owen** | Operations | Pro |
+| **Isaac** | Integration | Pro |
+| **Tommy** | Travel | Pro |
+| **Sloane** | Sales | Pro |
+| **Nadia** | UI/UX Design | Pro |
+| **Morgan** | Marketing | Pro |
+| **Alex** | API Crawling | Pro |
+| **Uma** | UX Research | Pro |
+| **Caleb** | CRM | Pro |
+| **Elena** | Estimation | Pro |
+| **Audrey** | Accounting | Pro |
+| **Logan** | Legal | Pro |
+| **Eva** | Executive Assistant | Pro |
+| **Parker** | Project Management | Pro |
+| **Aria** | Autonomous Research | Pro |
+| **Nate** | n8n Automation | Pro |
+| **MaxxiPro** | Roof Maxx Expert | Pro |
+| **AccuScope** | AccuLynx CRM Auditor | Pro |
+| **SmokeTestAgent** | QA Specialist | Pro |
+
+## How It Works
+
+1. **You talk to A.L.I.C.E.** — she's your single point of contact
+2. **A.L.I.C.E. routes to specialists** — "Build me an API" → Dylan; "Audit my security" → Selena
+3. **Specialists do the work** — using their domain tools and knowledge
+4. **A.L.I.C.E. synthesizes** — combines results into one clear answer
+
+Each specialist agent operates within their domain and flags when something falls outside their scope.
+
+## Model Configuration
+
+A.L.I.C.E. does **not** pick a model for you. It uses whatever your runtime already has configured.
+
+### Before Installing
+
+**Hermes Agent:**
+```bash
+hermes model
+# Follow the prompts to select a provider (e.g., OpenAI, Anthropic, Ollama, Nous Portal) and default model
+```
+
+**OpenClaw / NemoClaw:**
+```bash
+openclaw configure
+# Select "model" → choose your provider and default model
+```
+
+### Model Presets (OpenClaw/NemoClaw only)
+
+If your runtime already has a model configured, A.L.I.C.E. uses it automatically. The installer will not ask about models.
+
+If you need to override for OpenClaw/NemoClaw specifically, you can pass model hints before install:
+```bash
+# Set model in OpenClaw config directly
+openclaw configure --section model
+
+# Or configure API keys for your provider
+openclaw configure --section credentials
+```
 
 ## Install Options
 
 ```bash
-# Interactive install
+# Interactive install — auto-detects runtime and guides you through setup
 npx @robbiesrobotics/alice-agents
 
-# Non-interactive with defaults (detected model if available, otherwise Sonnet; Starter tier unless --tier pro)
+# Non-interactive (uses detected defaults)
 npx @robbiesrobotics/alice-agents --yes
 
-# Non-interactive Pro install with A.L.I.C.E. | Control Cloud enabled
-npx @robbiesrobotics/alice-agents --yes --tier pro --license-key YOUR_KEY --cloud --cloud-token YOUR_TOKEN
+# Pro tier
+npx @robbiesrobotics/alice-agents --yes --tier pro
 
-# Force the coding tool preference for this install
-npx @robbiesrobotics/alice-agents --yes --coding-tool codex
+# With A.L.I.C.E. | Control Cloud (Pro)
+npx @robbiesrobotics/alice-agents --yes --tier pro --license-key YOUR_KEY --cloud
 
-# Show help
-npx @robbiesrobotics/alice-agents --help
-```
+# Force a specific runtime (useful on machines with multiple runtimes)
+npx @robbiesrobotics/alice-agents --yes --runtime hermes
+npx @robbiesrobotics/alice-agents --yes --runtime openclaw
 
-### Install Modes
+# Reinstall over existing A.L.I.C.E.
+npx @robbiesrobotics/alice-agents --yes --force
 
-- **Fresh** — Replaces the agents section in `openclaw.json` (recommended for first install)
-- **Merge** — Adds A.L.I.C.E. agents alongside your existing agents
-- **Upgrade** — Updates product files (SOUL.md, AGENTS.md, etc.) without touching user customizations
+# Show all options
+npx @robbiesrobotics/alice-agents --help
+```
 
-### A.L.I.C.E. | Control Cloud
+## A.L.I.C.E. | Control Cloud
 
-If you're a Pro user with the cloud add-on, the installer can configure your local runtime for A.L.I.C.E. | Control in the same pass.
+A Pro add-on that connects your local runtime to the A.L.I.C.E. Control web dashboard. Enables telemetry, fleet visibility, and remote agent management.
 
-- Interactive install: choose `Pro`, validate your license, then enable the A.L.I.C.E. | Control Cloud add-on when prompted
-- Non-interactive install: pass `--tier pro --license-key YOUR_KEY --cloud`
-- Optional flags:
-  - `--cloud-token <token>` — access or ingest token for authenticated telemetry
-  - `--cloud-dashboard-url <url>` — defaults to `https://alice.av3.ai`
-  - `--cloud-ingest-url <url>` — defaults to `<dashboard-url>/api/v1/ingest`
+```bash
+npx @robbiesrobotics/alice-agents --yes --tier pro --license-key YOUR_KEY --cloud
+```
 
-The cloud config is stored in `~/.openclaw/.alice-mission-control.json`, and the bundled bridge plugin reads from that file so the setup remains portable across macOS, Linux, and Windows.
+Optional flags:
+- `--cloud-token <token>` — access or ingest token
+- `--cloud-dashboard-url <url>` — defaults to `https://alice.av3.ai`
+- `--cloud-ingest-url <url>` — defaults to `<dashboard-url>/api/v1/ingest`
 
-## Upgrade
+## Upgrading
 
-Re-run the installer and choose "Upgrade":
+```bash
+npx @robbiesrobotics/alice-agents --update
+```
 
+Or run interactively and select "Upgrade":
 ```bash
 npx @robbiesrobotics/alice-agents
-# Select: Upgrade
+# → Select: Upgrade
 ```
 
-This updates agent templates and config while preserving your PLAYBOOK.md, LEARNINGS.md, FEEDBACK.md, and memory/ directories.
+Upgrade preserves your `PLAYBOOK.md`, `LEARNINGS.md`, `FEEDBACK.md`, and `memory/` directories.
 
-## Uninstall
+## Uninstalling
 
 ```bash
 npx @robbiesrobotics/alice-agents --uninstall
 ```
 
-Removes A.L.I.C.E. agents from `openclaw.json` while preserving any non-ALICE agents. Creates a backup before making changes.
+Removes A.L.I.C.E. agents from your runtime config. Creates a backup first.
 
-## Maintainer Release Guard
+## Requirements
 
-Maintainer publishes are now gated.
+- **Hermes Agent** v0.10.0+ **or** **OpenClaw** / **NemoClaw** installed and configured
+- A default model set in your runtime (`hermes model` or `openclaw configure`)
+- Node.js 18+ (for the installer itself)
+- For local models: [Ollama](https://ollama.com) running locally
 
-- `npm publish` runs `prepublishOnly`
-- that hook runs tests, syntax checks, and `npm run release:check`
-- publish is blocked unless the package version has a matching git tag on `HEAD`
-- publish is blocked unless `landing/content/changelog.md` contains the matching version entry
-- publish is blocked unless `releases/vX.Y.Z.md` exists and is marked `Status: approved`
+## Architecture
 
-Use `releases/TEMPLATE.md` as the starting point for each release brief.
+A.L.I.C.E. is an orchestrator + specialist agent team. The orchestrator (A.L.I.C.E.) receives all requests, classifies intent, routes to the right specialist(s), and synthesizes results.
 
-## How It Works
+On **Hermes Agent:** A.L.I.C.E. and all specialists are installed as Hermes skills (`~/.hermes/skills/alice/`). Delegation is skill-to-skill via Hermes tool calling.
 
-1. **You talk to A.L.I.C.E.** — she's your single point of contact
-2. **A.L.I.C.E. routes to specialists** — "Build me an API" → Dylan (Development)
-3. **Specialists do the work** — using their domain-specific tools and expertise
-4. **A.L.I.C.E. synthesizes** — combines results and presents them to you
+On **OpenClaw / NemoClaw:** A.L.I.C.E. and specialists are installed as agent workspaces. Delegation uses `sessions_spawn`.
 
-Each agent has its own workspace with:
-- `SOUL.md` — personality and values
-- `AGENTS.md` — operating instructions
-- `PLAYBOOK.md` — learned patterns (evolves over time)
-- `LEARNINGS.md` — task log
-- `memory/` — persistent context between sessions
+Both architectures are functionally equivalent — the user experience is identical.
 
 ## Compatibility
 
-A.L.I.C.E. runs on any OpenClaw-compatible runtime:
-
-| Runtime | Status | Notes |
-|---------|--------|-------|
-| **NemoClaw** | ✅ Recommended | NVIDIA's secure distribution. Agents run inside OpenShell sandbox. [Get NemoClaw](https://nvidia.com/nemoclaw) |
-| **OpenClaw** | ✅ Fully supported | Lightweight option. [openclaw.ai](https://openclaw.ai) |
-
-### Why NemoClaw?
-
-NemoClaw is NVIDIA's open-source, enterprise-grade distribution of OpenClaw. It adds the OpenShell security sandbox — isolating what agents can access on your machine, enforcing policy-based guardrails, and keeping your data private. A.L.I.C.E. is NemoClaw-native as of v1.2.7: the installer defaults to NemoClaw, and agents display their sandbox status on startup.
-
-If you're already running NemoClaw, A.L.I.C.E. works out of the box — no extra config needed.
-
-## Requirements
-
-- [OpenClaw](https://openclaw.ai) or [NemoClaw](https://nemoclaw.com) installed and configured
-- Node.js 18+
-- At least one model configured in OpenClaw or NemoClaw — no specific API key required. A.L.I.C.E. uses whatever you already have set up.
+| Runtime | Agent Model | Config Location |
+|---------|-------------|----------------|
+| Hermes Agent | `~/.hermes/config.yaml` | `~/.hermes/skills/alice/` |
+| NemoClaw | `~/.nemoclaw/openclaw.json` | `~/.nemoclaw/workspace/` |
+| OpenClaw | `~/.openclaw/openclaw.json` | `~/.openclaw/workspace/` |
 
 ## License
 

package/bin/alice-install.mjs

@@ -33,6 +33,8 @@ if (flags.has('--help') || flags.has('-h')) {
     npx @robbiesrobotics/alice-agents --doctor     Run diagnostics on your A.L.I.C.E. install
     npx @robbiesrobotics/alice-agents --skills     Manage skills (install, remove, browse)
     npx @robbiesrobotics/alice-agents --cloud      Enable A.L.I.C.E. | Control Cloud during install
+    npx @robbiesrobotics/alice-agents --hermes-bridge  Enable Hermes agent bridge during install (OpenClaw+Hermes)
+    npx @robbiesrobotics/alice-agents --runtime <hermes|openclaw|nemoclaw>  Force runtime selection
     npx @robbiesrobotics/alice-agents --version    Show version
     npx @robbiesrobotics/alice-agents --help       Show this help
 
@@ -43,7 +45,10 @@ if (flags.has('--help') || flags.has('-h')) {
     --doctor      Run diagnostics and check install health
     --cloud       Enable A.L.I.C.E. | Control Cloud setup during install
     --no-cloud    Skip A.L.I.C.E. | Control Cloud setup during install
+    --hermes-bridge  Enable Hermes agent bridge (create hermes-agents.json, detect model)
     --tier <starter|pro>           Force the install tier
+    --runtime <hermes|openclaw|nemoclaw>  Force specific runtime
+    --force                        Force reinstall even if already installed
     --license-key <key>            Provide a Pro license key for automation
     --coding-tool <auto|claude|codex>  Override the preferred coding CLI
     --cloud-token <token>           A.L.I.C.E. | Control ingest/access token
@@ -58,30 +63,31 @@ if (flags.has('--help') || flags.has('-h')) {
   process.exit(0);
 }
 
+const baseOptions = {
+  tierOverride: getFlagValue('--tier'),
+  runtimeOverride: getFlagValue('--runtime'),
+  force: flags.has('--force'),
+  licenseKey: getFlagValue('--license-key'),
+  codingTool: getFlagValue('--coding-tool'),
+  cloudToken: getFlagValue('--cloud-token'),
+  cloudDashboardUrl: getFlagValue('--cloud-dashboard-url'),
+  cloudIngestUrl: getFlagValue('--cloud-ingest-url'),
+  cloudTeamId: getFlagValue('--cloud-team-id'),
+  cloudTeamSlug: getFlagValue('--cloud-team-slug'),
+  cloudTeamName: getFlagValue('--cloud-team-name'),
+  cloudTeamPlan: getFlagValue('--cloud-team-plan'),
+  hermesBridge: flags.has('--hermes-bridge'),
+};
+
 if (flags.has('--doctor')) {
   runDoctor().then((ok) => process.exit(ok ? 0 : 1)).catch((err) => {
     console.error('  ❌ Doctor failed:', err.message);
     process.exit(1);
   });
 } else if (flags.has('--update')) {
-  runInstall({
-    yes: true,
-    modeOverride: 'upgrade',
-    cloud: flags.has('--cloud') ? true : flags.has('--no-cloud') ? false : undefined,
-    tierOverride: getFlagValue('--tier'),
-    licenseKey: getFlagValue('--license-key'),
-    codingTool: getFlagValue('--coding-tool'),
-    cloudToken: getFlagValue('--cloud-token'),
-    cloudDashboardUrl: getFlagValue('--cloud-dashboard-url'),
-    cloudIngestUrl: getFlagValue('--cloud-ingest-url'),
-    cloudTeamId: getFlagValue('--cloud-team-id'),
-    cloudTeamSlug: getFlagValue('--cloud-team-slug'),
-    cloudTeamName: getFlagValue('--cloud-team-name'),
-    cloudTeamPlan: getFlagValue('--cloud-team-plan'),
-  }).then(() => process.exit(0)).catch((err) => {
-    console.error('  ❌ Update failed:', err.message);
-    process.exit(1);
-  });
+  runInstall({ yes: true, modeOverride: 'upgrade', cloud: flags.has('--cloud') ? true : flags.has('--no-cloud') ? false : undefined, ...baseOptions })
+    .then(() => process.exit(0))
+    .catch((err) => { console.error('  ❌ Upgrade failed:', err.message); process.exit(1); });
 } else if (flags.has('--skills')) {
   runSkillsManager().then(() => process.exit(0)).catch((err) => {
     console.error(`  ✗ Skills manager failed: ${err.message}`);
@@ -93,21 +99,7 @@ if (flags.has('--doctor')) {
     process.exit(1);
   });
 } else {
-  runInstall({
-    yes: flags.has('--yes'),
-    cloud: flags.has('--cloud') ? true : flags.has('--no-cloud') ? false : undefined,
-    tierOverride: getFlagValue('--tier'),
-    licenseKey: getFlagValue('--license-key'),
-    codingTool: getFlagValue('--coding-tool'),
-    cloudToken: getFlagValue('--cloud-token'),
-    cloudDashboardUrl: getFlagValue('--cloud-dashboard-url'),
-    cloudIngestUrl: getFlagValue('--cloud-ingest-url'),
-    cloudTeamId: getFlagValue('--cloud-team-id'),
-    cloudTeamSlug: getFlagValue('--cloud-team-slug'),
-    cloudTeamName: getFlagValue('--cloud-team-name'),
-    cloudTeamPlan: getFlagValue('--cloud-team-plan'),
-  }).then(() => process.exit(0)).catch((err) => {
-    console.error('  ❌ Install failed:', err.message);
-    process.exit(1);
-  });
+  runInstall({ yes: flags.has('--yes'), cloud: flags.has('--cloud') ? true : flags.has('--no-cloud') ? false : undefined, ...baseOptions })
+    .then(() => process.exit(0))
+    .catch((err) => { console.error('  ❌ Install failed:', err.message); process.exit(1); });
 }