multimodel-dev-os 2.0.1 → 2.8.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.ai/intelligence/README.md +14 -0
- package/.ai/intelligence/apply-log.schema.json +65 -0
- package/.ai/intelligence/feedback-log.example.jsonl +2 -0
- package/.ai/intelligence/feedback.schema.json +47 -0
- package/.ai/intelligence/improvement-proposal.schema.json +70 -0
- package/.ai/intelligence/learning-rules.example.md +18 -0
- package/.ai/intelligence/memory.schema.json +97 -0
- package/.ai/plugins/README.md +30 -0
- package/.ai/plugins/plugin.example.yaml +32 -0
- package/.ai/policies/approval-gates.md +35 -0
- package/.ai/policies/memory-policy.md +30 -0
- package/.ai/policies/self-improvement-policy.md +39 -0
- package/.ai/proposals/README.md +44 -0
- package/.ai/proposals/apply-operation.example.json +22 -0
- package/.ai/registries/capabilities.yaml +73 -0
- package/.ai/registries/tools.yaml +84 -0
- package/.ai/registries/workflows.yaml +217 -0
- package/.ai/schema/plugin.schema.json +56 -0
- package/README.md +116 -138
- package/assets/adapter-sync-flow.svg +84 -0
- package/assets/architecture-preview.svg +46 -31
- package/assets/onboarding-flow.svg +79 -0
- package/assets/social-preview.svg +1 -1
- package/assets/terminal-demo.svg +22 -23
- package/bin/multimodel-dev-os.js +3472 -7
- package/docs/.vitepress/config.js +46 -7
- package/docs/5-day-roadmap.md +9 -9
- package/docs/CLI.md +260 -34
- package/docs/adapter-sync.md +27 -0
- package/docs/adapters.md +16 -0
- package/docs/agent-handoff.md +40 -0
- package/docs/approved-proposal-apply.md +156 -0
- package/docs/architecture.md +31 -7
- package/docs/capability-registry.md +24 -0
- package/docs/comparison.md +72 -25
- package/docs/compatibility.md +2 -2
- package/docs/dashboard.md +105 -0
- package/docs/demo.md +23 -60
- package/docs/demos/adapter-sync.md +103 -0
- package/docs/demos/existing-repo-onboarding.md +125 -0
- package/docs/demos/index.md +91 -0
- package/docs/demos/multi-agent-handoff.md +88 -0
- package/docs/demos/release-check.md +109 -0
- package/docs/demos/safe-improvement-loop.md +119 -0
- package/docs/distribution.md +195 -0
- package/docs/faq.md +91 -24
- package/docs/feedback-learning.md +33 -0
- package/docs/future-proof-architecture.md +22 -0
- package/docs/hash-compressed-memory.md +72 -0
- package/docs/improvement-proposals.md +70 -0
- package/docs/index.md +192 -81
- package/docs/installers.md +18 -4
- package/docs/launch-kit.md +97 -49
- package/docs/learning-rules.md +36 -0
- package/docs/npm-publishing.md +6 -6
- package/docs/plugin-authoring.md +99 -0
- package/docs/plugin-hooks.md +80 -0
- package/docs/public/assets/adapter-sync-flow.svg +84 -0
- package/docs/public/assets/onboarding-flow.svg +79 -0
- package/docs/public/llms-full.txt +47 -4
- package/docs/public/llms.txt +55 -2
- package/docs/public/sitemap.xml +85 -0
- package/docs/quickstart.md +82 -22
- package/docs/real-repo-onboarding.md +27 -0
- package/docs/repository-command-center.md +68 -0
- package/docs/self-improving-codebase.md +46 -0
- package/docs/template-recommendation.md +22 -0
- package/docs/templates-guide.md +11 -0
- package/docs/tool-registry.md +21 -0
- package/docs/tui-safety.md +59 -0
- package/docs/use-cases.md +21 -0
- package/docs/v2-roadmap.md +78 -71
- package/docs/workflow-orchestration.md +62 -0
- package/examples/adapter-sync/README.md +45 -0
- package/examples/command-center/README.md +59 -0
- package/examples/real-repo-onboarding/README.md +53 -0
- package/examples/safe-improvement-loop/README.md +48 -0
- package/package.json +1 -1
- package/scripts/install.ps1 +1 -1
- package/scripts/install.sh +1 -1
- package/scripts/verify.js +107 -3
package/docs/quickstart.md
CHANGED
|
@@ -1,31 +1,33 @@
|
|
|
1
|
-
# Quickstart Guide
|
|
1
|
+
# Quickstart Guide
|
|
2
2
|
|
|
3
|
-
Get
|
|
3
|
+
Get MultiModel Dev OS integrated into your codebase in **under 30 seconds** to synchronize your multi-agent developer workflows.
|
|
4
4
|
|
|
5
|
-
> **Use when**: Setting up a new repository or aligning multiple AI tools (
|
|
5
|
+
> **Use when**: Setting up a new repository or aligning multiple AI tools (Cursor, Claude Code, Gemini, Codex, Antigravity, VS Code) to prevent instruction drift and prompt token bloat.
|
|
6
6
|
|
|
7
7
|
---
|
|
8
8
|
|
|
9
|
-
## Option A: NPX Scaffolding (
|
|
9
|
+
## Option A: NPX Scaffolding (Recommended)
|
|
10
10
|
|
|
11
|
-
Initialize any project instantly using
|
|
11
|
+
Initialize any project instantly using the public npm registry:
|
|
12
12
|
|
|
13
13
|
```bash
|
|
14
|
-
# Standard
|
|
14
|
+
# Standard initialization in the current directory
|
|
15
15
|
npx multimodel-dev-os@latest init
|
|
16
16
|
|
|
17
|
-
#
|
|
17
|
+
# Pick a stack template and inject specific adapters
|
|
18
18
|
npx multimodel-dev-os@latest init --template nextjs-saas --adapter cursor --adapter claude
|
|
19
19
|
|
|
20
|
-
# Run a dry-run preview before
|
|
20
|
+
# Run a dry-run preview before writing any files
|
|
21
21
|
npx multimodel-dev-os@latest init --dry-run
|
|
22
22
|
```
|
|
23
23
|
|
|
24
|
+
**Available templates:** `nextjs-saas` · `wordpress-site` · `ecommerce-store` · `seo-landing-page` · `expo-react-native-android` · `general-app`
|
|
25
|
+
|
|
24
26
|
---
|
|
25
27
|
|
|
26
28
|
## Option B: Fallback One-Line Scripts
|
|
27
29
|
|
|
28
|
-
If you
|
|
30
|
+
If you prefer installation scripts:
|
|
29
31
|
|
|
30
32
|
**macOS / Linux / WSL (bash):**
|
|
31
33
|
```bash
|
|
@@ -41,17 +43,17 @@ irm https://raw.githubusercontent.com/rizvee/multimodel-dev-os/main/scripts/inst
|
|
|
41
43
|
|
|
42
44
|
## Option C: Caveman Mode (Minimal Tokens)
|
|
43
45
|
|
|
44
|
-
|
|
46
|
+
Cut prompt token overhead by **~79%** with compressed shorthand declarations. Best for tight API budgets or smaller models:
|
|
45
47
|
|
|
46
48
|
```bash
|
|
47
|
-
|
|
49
|
+
npx multimodel-dev-os@latest init --caveman
|
|
48
50
|
```
|
|
49
51
|
|
|
50
52
|
---
|
|
51
53
|
|
|
52
|
-
## Option D:
|
|
54
|
+
## Option D: Target a Specific Directory
|
|
53
55
|
|
|
54
|
-
|
|
56
|
+
Initialize a target project path directly:
|
|
55
57
|
|
|
56
58
|
```bash
|
|
57
59
|
npx multimodel-dev-os@latest init --target /path/to/your-project --template nextjs-saas --adapter cursor
|
|
@@ -59,28 +61,86 @@ npx multimodel-dev-os@latest init --target /path/to/your-project --template next
|
|
|
59
61
|
|
|
60
62
|
---
|
|
61
63
|
|
|
64
|
+
## Option E: Onboard an Existing Repository
|
|
65
|
+
|
|
66
|
+
Already have a project with code in it? Use the onboarding system to safely analyze your repo and bootstrap configs without breaking anything:
|
|
67
|
+
|
|
68
|
+
```bash
|
|
69
|
+
# Step 1: Analyze your project structure
|
|
70
|
+
npx multimodel-dev-os@latest onboard analyze
|
|
71
|
+
|
|
72
|
+
# Step 2: Get template and adapter recommendations
|
|
73
|
+
npx multimodel-dev-os@latest onboard recommend
|
|
74
|
+
|
|
75
|
+
# Step 3: Generate an onboarding plan
|
|
76
|
+
npx multimodel-dev-os@latest onboard plan
|
|
77
|
+
|
|
78
|
+
# Step 4: Apply the plan (creates backups automatically)
|
|
79
|
+
npx multimodel-dev-os@latest onboard apply --approved
|
|
80
|
+
|
|
81
|
+
# Step 5: Check onboarding status
|
|
82
|
+
npx multimodel-dev-os@latest onboard status
|
|
83
|
+
```
|
|
84
|
+
|
|
85
|
+
---
|
|
86
|
+
|
|
87
|
+
## Option F: Launch the Interactive Dashboard
|
|
88
|
+
|
|
89
|
+
Launch the TUI command center to manage onboarding, adapter sync, memory building, feedback collection, proposals, and quality checks in an interactive menu:
|
|
90
|
+
|
|
91
|
+
```bash
|
|
92
|
+
npx multimodel-dev-os@latest dashboard
|
|
93
|
+
```
|
|
94
|
+
|
|
95
|
+
---
|
|
96
|
+
|
|
62
97
|
## After Install
|
|
63
98
|
|
|
64
|
-
1. **Edit `AGENTS.md`** — fill in your project name, stack, and build commands
|
|
99
|
+
1. **Edit `AGENTS.md`** — fill in your project name, stack, and build commands.
|
|
65
100
|
2. **Edit `.ai/config.yaml`** — enable adapters for your tools.
|
|
66
|
-
3. **
|
|
101
|
+
3. **Sync adapter rule files** to your project root automatically:
|
|
102
|
+
```bash
|
|
103
|
+
npx multimodel-dev-os@latest adapter sync all --approved
|
|
104
|
+
```
|
|
105
|
+
Or copy manually if you prefer:
|
|
67
106
|
- Cursor: `cp adapters/cursor/.cursorrules .cursorrules`
|
|
68
107
|
- Claude: `cp adapters/claude/CLAUDE.md CLAUDE.md`
|
|
69
108
|
- VS Code: `cp -r adapters/vscode/.vscode/ .vscode/`
|
|
70
|
-
4. **
|
|
109
|
+
4. **Build your codebase memory index:**
|
|
110
|
+
```bash
|
|
111
|
+
npx multimodel-dev-os@latest memory build
|
|
112
|
+
```
|
|
113
|
+
5. **Start coding** — your AI coding agents will read the shared configuration instantly.
|
|
71
114
|
|
|
72
115
|
---
|
|
73
116
|
|
|
74
|
-
##
|
|
117
|
+
## Try the Demo Workflows
|
|
118
|
+
|
|
119
|
+
To see MultiModel Dev OS in action in under 2 minutes, try one of our guided, copy-paste demo workflows:
|
|
120
|
+
- 🚀 **[Safe Repo Onboarding](/demos/existing-repo-onboarding)** — Safely onboard an existing project without breaking changes.
|
|
121
|
+
- 🔄 **[Universal Adapter Sync](/demos/adapter-sync)** — Write rules once in `AGENTS.md` and sync Cursor, Claude, and Gemini.
|
|
122
|
+
- 🤝 **[Multi-Agent Handoff](/demos/multi-agent-handoff)** — Smoothly hand off session context between terminal and editor agents.
|
|
123
|
+
- 🔁 **[Safe Improvement Loop](/demos/safe-improvement-loop)** — Capture user feedback and run safe, audited codebase upgrades.
|
|
124
|
+
- 🩺 **[Release Verification](/demos/release-check)** — Run the final pre-flight checks before releasing a package.
|
|
75
125
|
|
|
76
|
-
|
|
126
|
+
Or browse them all on the [Demos Hub](/demos/).
|
|
127
|
+
|
|
128
|
+
---
|
|
129
|
+
|
|
130
|
+
## Verify & Diagnose
|
|
77
131
|
|
|
78
132
|
```bash
|
|
79
133
|
# Strict directory schema validation
|
|
80
|
-
npx multimodel-dev-os validate
|
|
134
|
+
npx multimodel-dev-os@latest validate
|
|
135
|
+
|
|
136
|
+
# Advisory workspace compatibility audit
|
|
137
|
+
npx multimodel-dev-os@latest doctor
|
|
138
|
+
|
|
139
|
+
# Codebase structure scanner
|
|
140
|
+
npx multimodel-dev-os@latest scan
|
|
81
141
|
|
|
82
|
-
#
|
|
83
|
-
npx multimodel-dev-os
|
|
142
|
+
# Compact status dashboard
|
|
143
|
+
npx multimodel-dev-os@latest status
|
|
84
144
|
```
|
|
85
145
|
|
|
86
|
-
Explore
|
|
146
|
+
Explore the full **[CLI Command Reference](/CLI)**, **[Stable Protocol Specification](/stable-protocol)**, or **[Upgrade & Migration Guide](/migration-guide)** for details.
|
|
@@ -0,0 +1,27 @@
|
|
|
1
|
+
# Onboarding Existing Repositories
|
|
2
|
+
|
|
3
|
+
MultiModel Dev OS provides a safe, step-by-step onboarding protocol to configure coding agents in any existing codebase without modifying application source files.
|
|
4
|
+
|
|
5
|
+
## Onboarding Commands
|
|
6
|
+
|
|
7
|
+
To onboard an existing repository, run the following commands:
|
|
8
|
+
|
|
9
|
+
```bash
|
|
10
|
+
# 1. Analyze the project structure, language, framework, and security signals
|
|
11
|
+
npx multimodel-dev-os onboard analyze --target .
|
|
12
|
+
|
|
13
|
+
# 2. Output template recommendations and suggested next steps
|
|
14
|
+
npx multimodel-dev-os onboard recommend --target .
|
|
15
|
+
|
|
16
|
+
# 3. Generate a structured onboarding plan and markdown report
|
|
17
|
+
npx multimodel-dev-os onboard plan --target .
|
|
18
|
+
|
|
19
|
+
# 4. Safely apply the onboarding plan to generate AI Dev OS scaffolding
|
|
20
|
+
npx multimodel-dev-os onboard apply --target . --approved
|
|
21
|
+
```
|
|
22
|
+
|
|
23
|
+
## Security & Safety Guarantees
|
|
24
|
+
|
|
25
|
+
* **Read-only Commands**: `analyze`, `recommend`, and `plan` are read-only and make no modifications.
|
|
26
|
+
* **No Source Code Mutation**: `apply` only creates/updates MultiModel Dev OS configuration files (`.ai/`, `AGENTS.md`, `MEMORY.md`, etc.). It never touches your application source code.
|
|
27
|
+
* **Backup Support**: If a file already exists, the onboard engine skips it. To overwrite, use `--force`, which creates a `.bak` backup file first.
|
|
@@ -0,0 +1,68 @@
|
|
|
1
|
+
# Repository Command Center Guide
|
|
2
|
+
|
|
3
|
+
The **Repository Command Center** introduced in `v2.5.0` provides a unified, read-only operational dashboard summarizing the entire repository intelligence status of a MultiModel Dev OS workspace.
|
|
4
|
+
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
## 1. Using the `status` Command
|
|
8
|
+
|
|
9
|
+
The `status` command allows developers and AI agents to instantly check project health and progress.
|
|
10
|
+
|
|
11
|
+
```bash
|
|
12
|
+
npx multimodel-dev-os status
|
|
13
|
+
```
|
|
14
|
+
|
|
15
|
+
Or target a specific subdirectory:
|
|
16
|
+
|
|
17
|
+
```bash
|
|
18
|
+
npx multimodel-dev-os status --target ./some-subproject
|
|
19
|
+
```
|
|
20
|
+
|
|
21
|
+
### Dashboard View
|
|
22
|
+
|
|
23
|
+
Running `status` prints a formatted, color-coded status summary:
|
|
24
|
+
|
|
25
|
+
1. **Project Info**: Displays the package name and package version from the root `package.json`.
|
|
26
|
+
2. **Framework & Dependency Signals**: Summarizes detected languages, framework engines (e.g. Next.js, Express, React), and package lockfiles (e.g. `package-lock.json`).
|
|
27
|
+
3. **Memory State**:
|
|
28
|
+
* `CURRENT` (Green): Memory index is fully built and matches the filesystem.
|
|
29
|
+
* `STALE` (Yellow): Modifications have been made. Lists number of added, removed, or modified files.
|
|
30
|
+
* `MISSING` (Red): No memory files have been built yet.
|
|
31
|
+
4. **Feedback Loop & Rules**: Displays total count of developer feedback entries and whether compiled learning rules are active.
|
|
32
|
+
5. **Improvement Proposals**: Lists total optimization proposals, categorized by approval status (`pending`, `approved`, `rejected`).
|
|
33
|
+
6. **Apply Audit Log**: Counts applied proposal runs.
|
|
34
|
+
7. **Next Recommended Command**: Offers dynamic advice on what to execute next (e.g., refreshing memory, review proposals, running health checks).
|
|
35
|
+
|
|
36
|
+
---
|
|
37
|
+
|
|
38
|
+
## 2. Benefits for Multi-Agent Workflows
|
|
39
|
+
|
|
40
|
+
* **Context Discovery**: Instantly summarizes active templates, environment structures, and boundaries.
|
|
41
|
+
* **Zero-Dependency Parsing**: Fast initialization checks that don't load external libraries.
|
|
42
|
+
* **Actionable Next Steps**: Guides developers and AI models to the correct command sequence, preventing execution drift.
|
|
43
|
+
|
|
44
|
+
## 3. Onboarding Status
|
|
45
|
+
|
|
46
|
+
For uninitialized repositories, you can inspect the onboarding status directly:
|
|
47
|
+
|
|
48
|
+
```bash
|
|
49
|
+
npx multimodel-dev-os onboard status
|
|
50
|
+
```
|
|
51
|
+
|
|
52
|
+
This prints a checklist of crucial MultiModel Dev OS files and an overall completeness percentage score.
|
|
53
|
+
|
|
54
|
+
---
|
|
55
|
+
|
|
56
|
+
## 4. Interactive TUI Command Center
|
|
57
|
+
|
|
58
|
+
Starting in `v2.8.0`, MultiModel Dev OS introduces a full interactive terminal user interface (TUI) command center that wraps all system diagnostics, sync tools, and memory engines.
|
|
59
|
+
|
|
60
|
+
To launch the interactive dashboard, run:
|
|
61
|
+
|
|
62
|
+
```bash
|
|
63
|
+
npx multimodel-dev-os dashboard
|
|
64
|
+
# or:
|
|
65
|
+
npx multimodel-dev-os ui
|
|
66
|
+
```
|
|
67
|
+
|
|
68
|
+
For detailed controls, menu hierarchies, and automated CI safety fallbacks, refer to the [Interactive TUI Dashboard Guide](file:///f:/multimodel-dev-os/docs/dashboard.md).
|
|
@@ -0,0 +1,46 @@
|
|
|
1
|
+
# Self-Improving Codebase Operations
|
|
2
|
+
|
|
3
|
+
MultiModel Dev OS supports codebase refactoring, styling, and structural optimizations through a secure, proposal-based self-improvement pipeline.
|
|
4
|
+
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
## 1. Safety Principles
|
|
8
|
+
|
|
9
|
+
To guarantee repository integrity and prevent unintended changes, all self-improvement operations adhere to three core rules:
|
|
10
|
+
1. **Write-Protection**: The CLI binary, security guards, workflows, and core policies are read-only and cannot be modified by model proposals.
|
|
11
|
+
2. **Human-in-the-Loop (HITL)**: Models cannot write code directly to the workspace without explicit user verification.
|
|
12
|
+
3. **Diff-Validated Execution**: Every modification must pass local verification tests, with manual commit approval.
|
|
13
|
+
|
|
14
|
+
---
|
|
15
|
+
|
|
16
|
+
## 2. Proposal-Review Cycle
|
|
17
|
+
|
|
18
|
+
The execution path follows two stages:
|
|
19
|
+
|
|
20
|
+
### Stage A: Proposal Generation
|
|
21
|
+
Before compiling a proposal, the agent scans the codebase using `mmdo scan` and builds the hash-compressed memory index via `mmdo memory build` to ensure the proposal is context-aware and token-efficient. The agent then identifies an optimization and saves a structured document under `.ai/proposals/proposal-YYYYMMDD-HHMMSS.md`. The proposal specifies the rationale, affected files list, risk level, test command, and rollback steps.
|
|
22
|
+
|
|
23
|
+
### Stage B: Human Review
|
|
24
|
+
The developer runs the review CLI command:
|
|
25
|
+
```bash
|
|
26
|
+
npx multimodel-dev-os improve review
|
|
27
|
+
```
|
|
28
|
+
This shows the proposed changes, affected files, risk levels, and verify commands.
|
|
29
|
+
|
|
30
|
+
To view status:
|
|
31
|
+
```bash
|
|
32
|
+
npx multimodel-dev-os improve status
|
|
33
|
+
```
|
|
34
|
+
|
|
35
|
+
*Note: Starting in v2.4.1, automated proposal application is supported via the `improve apply` command, provided that a valid operations JSON block exists in the proposal file and the developer passes the `--approved` flag. Vague proposals or those targeting protected files/paths must still be executed manually.*
|
|
36
|
+
|
|
37
|
+
For complete automated application details, please refer to the [Approved Proposal Application Guide](/approved-proposal-apply).
|
|
38
|
+
|
|
39
|
+
---
|
|
40
|
+
|
|
41
|
+
## Command Center Integration
|
|
42
|
+
|
|
43
|
+
The [Repository Command Center](repository-command-center.md) serves as the unified dashboard for the self-improvement loop. Running `mmdo status` displays the counts of pending/approved/rejected proposals, memory state, and feedback logs. Developers can orchestrate the review cycle safely by executing the `proposal-review` workflow:
|
|
44
|
+
```bash
|
|
45
|
+
npx multimodel-dev-os workflow run proposal-review
|
|
46
|
+
```
|
|
@@ -0,0 +1,22 @@
|
|
|
1
|
+
# Template Recommendation Engine
|
|
2
|
+
|
|
3
|
+
The recommendation engine matches existing codebases to the best-fit template profile.
|
|
4
|
+
|
|
5
|
+
## Heuristics & Confidence Scores
|
|
6
|
+
|
|
7
|
+
The engine assigns weights to file signatures and package dependencies:
|
|
8
|
+
|
|
9
|
+
* **Next.js SaaS (`nextjs-saas`)**: Matches if `next.config.js` or `next.config.ts` is present (Weight: +0.6) or `package.json` contains `next` dependency (Weight: +0.4).
|
|
10
|
+
* **Android Expo Mobile (`expo-react-native-android`)**: Matches if `app.json` or `eas.json` is present (Weight: +0.8).
|
|
11
|
+
* **WordPress Theme/Plugin (`wordpress-theme-plugin`)**: Matches if `wp-content` directory or header-decorated PHP style files are found (Weight: +0.8).
|
|
12
|
+
* **E-Commerce Store (`ecommerce-store`)**: Matches if Stripe, Shopify, or checkout-related dependencies are present (Weight: +0.4).
|
|
13
|
+
* **SEO Landing Page (`seo-landing-page`)**: Matches if raw HTML files + Tailwind configs exist without framework elements (Weight: +0.5).
|
|
14
|
+
* **General App (`general-app`)**: Serves as the fallback profile if no other template achieves a confidence score greater than 0.4.
|
|
15
|
+
|
|
16
|
+
## Command
|
|
17
|
+
|
|
18
|
+
To check template recommendations for your project:
|
|
19
|
+
|
|
20
|
+
```bash
|
|
21
|
+
npx multimodel-dev-os onboard recommend --target .
|
|
22
|
+
```
|
package/docs/templates-guide.md
CHANGED
|
@@ -69,3 +69,14 @@ npx multimodel-dev-os init --target ../my-new-app --template nextjs-saas
|
|
|
69
69
|
## Template Quality Assurance
|
|
70
70
|
|
|
71
71
|
To ensure all contributed templates meet our rigorous quality standards before packaging, refer to our [Template Quality Assurance Blueprint](/template-qa).
|
|
72
|
+
|
|
73
|
+
## Template Recommendations for Existing Repos
|
|
74
|
+
|
|
75
|
+
If you are onboarding an existing repository and unsure which template profile to use, the recommendation engine can automatically analyze your project files:
|
|
76
|
+
|
|
77
|
+
```bash
|
|
78
|
+
# Analyze framework and package signals, then output recommendation
|
|
79
|
+
npx multimodel-dev-os onboard recommend --target .
|
|
80
|
+
```
|
|
81
|
+
|
|
82
|
+
For detailed heuristics and confidence weights, refer to the [Template Recommendation Engine docs](/template-recommendation).
|
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
# Tool & Protocol Registry
|
|
2
|
+
|
|
3
|
+
The **Tool Registry** governs connection protocols, interfaces, and file associations for developer environments and AI agents.
|
|
4
|
+
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
## 1. Tool Mappings
|
|
8
|
+
|
|
9
|
+
Tools configured in `.ai/registries/tools.yaml` are mapped by:
|
|
10
|
+
* **Interface Type**: `editor` (IDE plugins), `terminal` (CLI clients), or `assistant` (chat helpers).
|
|
11
|
+
* **Protocols**: Connection formats such as `stdin-stdout`, `mcp-jsonrpc` (Model Context Protocol), or custom config parameters.
|
|
12
|
+
* **Registry Keys**: Associated configuration files (e.g. `.cursorrules` for Cursor, `CLAUDE.md` for Claude Code).
|
|
13
|
+
|
|
14
|
+
---
|
|
15
|
+
|
|
16
|
+
## 2. Model Context Protocol (MCP) Integration
|
|
17
|
+
|
|
18
|
+
MultiModel Dev OS leverages MCP to expose local resources to models:
|
|
19
|
+
1. **Server Bindings**: Developers declare local MCP servers (e.g. gcloud SDK MCP, Chrome DevTools MCP) under the tools registry.
|
|
20
|
+
2. **Capability Matching**: The Capability Registry checks if the selected model has `mcp-compliance: 1.0` before attempting native bindings.
|
|
21
|
+
3. **Unified Tool Access**: Exposes filesystem, browser automation, and deployment commands directly to agent execution loops securely.
|
|
@@ -0,0 +1,59 @@
|
|
|
1
|
+
# TUI & Plugin Safety Model
|
|
2
|
+
|
|
3
|
+
MultiModel Dev OS enforces a sandboxed, low-risk execution model for both the Interactive TUI Dashboard and the Declarative Plugin System. This document outlines the security architecture and constraints that protect your development environment.
|
|
4
|
+
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
## Interactive Dashboard Safety
|
|
8
|
+
|
|
9
|
+
The `dashboard`/`ui` command acts as a wrapper around the existing CLI commands.
|
|
10
|
+
|
|
11
|
+
### 1. Zero Arbitrary Code Execution
|
|
12
|
+
The TUI engine is written entirely using Node.js's native `readline` module. It does not pull in any external third-party interactive libraries or prompt managers at runtime. This guarantees that no untrusted node package codes are executed during dashboard rendering.
|
|
13
|
+
|
|
14
|
+
### 2. Explicit Approval Gates
|
|
15
|
+
Any operation that alters the workspace directory (such as adapter synchronization or onboarding template setup) is configured to run in **Dry-Run mode** by default. To apply modifications, the underlying command must be run directly from the shell with the explicit `--approved` and `--force` flags.
|
|
16
|
+
|
|
17
|
+
### 3. Equivalent Command Visibility
|
|
18
|
+
Before executing any action from the menu, the dashboard prints the exact CLI command in bold text. This removes hidden actions and teaches the developer the underlying tooling mechanics.
|
|
19
|
+
|
|
20
|
+
---
|
|
21
|
+
|
|
22
|
+
## Declarative Plugin Sandboxing
|
|
23
|
+
|
|
24
|
+
The plugin system allows community-contributed extensions while ensuring that they cannot infect the host machine or repository code.
|
|
25
|
+
|
|
26
|
+
### 1. Strictly Config-Based (YAML)
|
|
27
|
+
Plugins in MMDO cannot declare execution steps, binaries, or javascript files. They contain only metadata, workflow steps (mapping to pre-approved MMDO commands), and file paths to copy.
|
|
28
|
+
|
|
29
|
+
### 2. Path Whitelisting (Write Containment)
|
|
30
|
+
The installer restricts all plugin file copies to specific directories inside the workspace. The whitelisted paths are:
|
|
31
|
+
* `.ai/plugins/`
|
|
32
|
+
* `.ai/registries/`
|
|
33
|
+
* `.ai/templates/`
|
|
34
|
+
* `.ai/skills/`
|
|
35
|
+
* `.ai/checks/`
|
|
36
|
+
* `.ai/prompts/`
|
|
37
|
+
* `.ai/adapters/`
|
|
38
|
+
|
|
39
|
+
If a plugin manifest attempts to write to paths outside of these folders (such as `src/`, `lib/`, or `tests/`), the validator immediately throws an error and aborts the installation.
|
|
40
|
+
|
|
41
|
+
### 3. Blacklist Enforcement
|
|
42
|
+
The system hard-blocks write operations to critical files or directories to prevent credentials exposure or project hijack:
|
|
43
|
+
* `.env` and `.env.*` (secrets containment)
|
|
44
|
+
* `.git/` (git history protection)
|
|
45
|
+
* `.npmrc` (npm authentication protection)
|
|
46
|
+
* `node_modules/` (dependency safety)
|
|
47
|
+
* `package.json` and `package-lock.json` (preventing unauthorized package installations)
|
|
48
|
+
|
|
49
|
+
### 4. Overwrite & Conflict Backups
|
|
50
|
+
If a plugin attempts to copy files over existing workspace files:
|
|
51
|
+
* Without `--force`, the installation aborts with conflict warnings.
|
|
52
|
+
* With `--force`, the installer automatically creates backup copies (`<filename>.bak`) of the target files before overwriting them.
|
|
53
|
+
|
|
54
|
+
---
|
|
55
|
+
|
|
56
|
+
## Network & Package Quarantine
|
|
57
|
+
|
|
58
|
+
* **No Network Calls**: Neither the dashboard nor the plugin engine executes network queries (e.g. `curl`, `fetch`, or remote API calls). All plugin checks and file copy actions are performed 100% offline.
|
|
59
|
+
* **No Package Downloads**: Plugins cannot run `npm install` or call other package managers.
|
package/docs/use-cases.md
CHANGED
|
@@ -36,4 +36,25 @@ The universal fallback layout for Python, Go, Rust, or general backend setups.
|
|
|
36
36
|
|
|
37
37
|
---
|
|
38
38
|
|
|
39
|
+
## Workflow-Focused Use Cases
|
|
40
|
+
|
|
41
|
+
Beyond specific technology stacks, MultiModel Dev OS addresses operational friction in modern development environments:
|
|
42
|
+
|
|
43
|
+
### A. Onboarding Legacy Repositories
|
|
44
|
+
- **The Challenge:** Joining a massive legacy project with sparse docs. Developers spend days learning layout conventions, and AI assistants generate hallucinated file structures.
|
|
45
|
+
- **The Solution:** Run `onboard analyze` to automatically map stack layers, followed by `init` to bootstrap structured rules. Developers instantly get a baseline `AGENTS.md` and `.ai/` directory that teaches incoming models exactly where files belong and how the project compiles.
|
|
46
|
+
- **Demo:** Check out the **[Safe Onboarding Workflow](/demos/existing-repo-onboarding)**.
|
|
47
|
+
|
|
48
|
+
### B. Cross-Tool Collaboration
|
|
49
|
+
- **The Challenge:** Using Cursor for quick autocomplete while using Claude Code in the terminal to execute builds. Updates to rules in one tool aren't picked up by the other, leading to conflicting edits.
|
|
50
|
+
- **The Solution:** Standardize rules in `AGENTS.md` and use `adapter sync` to write rules to all tool configurations in a single run. Both terminal agents and editor assistants stay aligned.
|
|
51
|
+
- **Demo:** Check out the **[Universal Adapter Sync Workflow](/demos/adapter-sync)**.
|
|
52
|
+
|
|
53
|
+
### C. Audited Self-Improvement Loops
|
|
54
|
+
- **The Challenge:** Allowing autonomous agents to apply codebase improvements without a solid verification pipeline can lead to broken builds or silent data loss.
|
|
55
|
+
- **The Solution:** Establish a feedback loop where models propose modifications as patch proposals. MultiModel Dev OS enforces Human-In-The-Loop (HITL) approval flags alongside automated `validate` and `doctor` checks before applying edits.
|
|
56
|
+
- **Demo:** Check out the **[Safe Improvement Loop Workflow](/demos/safe-improvement-loop)**.
|
|
57
|
+
|
|
58
|
+
---
|
|
59
|
+
|
|
39
60
|
For a complete breakdown of all files, scaffolding structures, and custom developer skills included in each profile, refer to the [Templates Guide](templates-guide.md).
|
package/docs/v2-roadmap.md
CHANGED
|
@@ -1,95 +1,102 @@
|
|
|
1
|
-
# MultiModel Dev OS — v2.
|
|
1
|
+
# MultiModel Dev OS — Roadmap: v2.x → v3.0
|
|
2
2
|
|
|
3
|
-
This document outlines the development path,
|
|
3
|
+
This document outlines the development path, completed milestones, and future plans for MultiModel Dev OS.
|
|
4
4
|
|
|
5
5
|
---
|
|
6
6
|
|
|
7
|
-
## 1.
|
|
8
|
-
|
|
9
|
-
The primary goal of the **v2.0.0 release** is to promote the experimental features introduced in `v1.2.0` (Template Galaxy, Model Compatibility Layer, and Android Expo template) into officially frozen, production-grade core components, and resume stable package publication to the public npm registry.
|
|
7
|
+
## 1. Current Status
|
|
10
8
|
|
|
11
9
|
> [!IMPORTANT]
|
|
12
|
-
> **v2.0
|
|
10
|
+
> **v2.7.0 is the active stable release** on the public npm registry. All features below marked ✅ are shipped and production-ready.
|
|
13
11
|
|
|
14
12
|
---
|
|
15
13
|
|
|
16
|
-
## 2.
|
|
17
|
-
|
|
18
|
-
###
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
22
|
-
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
|
|
29
|
-
|
|
30
|
-
|
|
31
|
-
|
|
32
|
-
|
|
33
|
-
|
|
34
|
-
|
|
35
|
-
|
|
36
|
-
|
|
37
|
-
|
|
38
|
-
|
|
39
|
-
|
|
40
|
-
|
|
41
|
-
|
|
42
|
-
|
|
43
|
-
|
|
44
|
-
|
|
45
|
-
|
|
46
|
-
|
|
47
|
-
|
|
14
|
+
## 2. Completed Milestones
|
|
15
|
+
|
|
16
|
+
### v2.0.0 — Template Galaxy & Model Registry ✅
|
|
17
|
+
- Standardized model registries under `.ai/models/` (registry, providers, routing presets, local models)
|
|
18
|
+
- Adapter registry expansion under `.ai/adapters/registry.yaml`
|
|
19
|
+
- Android Expo mobile template (`examples/expo-react-native-android/`)
|
|
20
|
+
- CLI registry subcommands: `models`, `show-model`, `providers`, `route-model`, `adapters`, `show-adapter`, `skills`, `show-skill`
|
|
21
|
+
- Stable npm publication resumed
|
|
22
|
+
|
|
23
|
+
### v2.2.0 — Codebase Scanner & Memory Engine ✅
|
|
24
|
+
- `scan` command to inspect frameworks, package managers, and AI Dev OS files
|
|
25
|
+
- `memory build`, `memory refresh`, and `memory diff` commands
|
|
26
|
+
- Hash-compressed memory indexing to `.ai/intelligence/memory.hash.json`
|
|
27
|
+
- Secret-safety exclusions for `.env`, `.npmrc`, `.keystore` files
|
|
28
|
+
|
|
29
|
+
### v2.3.0 — Feedback Learning & Proposal Engine ✅
|
|
30
|
+
- `feedback add`, `feedback list`, `feedback summarize` commands
|
|
31
|
+
- `improve propose`, `improve review`, `improve status` commands
|
|
32
|
+
- Read-only proposal drafting with safety gates
|
|
33
|
+
- Feedback logs and learning rules compilation
|
|
34
|
+
|
|
35
|
+
### v2.4.0 — Approved Proposal Application Engine ✅
|
|
36
|
+
- `improve validate`, `improve diff`, `improve apply` subcommands
|
|
37
|
+
- 12 strict safety gates including path boundary, protected paths, idempotency
|
|
38
|
+
- Applied Proposals Audit Log (`apply-log.jsonl`) with SHA-256 hashing
|
|
39
|
+
- Deterministic operations: `create_file`, `append_line`, `replace_text`
|
|
40
|
+
|
|
41
|
+
### v2.5.0 — Repository Intelligence Command Center ✅
|
|
42
|
+
- `status` command — compact operational dashboard
|
|
43
|
+
- `workflow list`, `show`, `plan`, `run` — multi-agent workflow orchestration
|
|
44
|
+
- `handoff build`, `handoff show` — token-compressed session context
|
|
45
|
+
- Safe execution boundaries — no destructive operations from workflows
|
|
46
|
+
|
|
47
|
+
### v2.6.0 — Real-Repo Onboarding & Adapter Sync ✅
|
|
48
|
+
- `onboard analyze`, `recommend`, `plan`, `apply`, `status` — existing repo onboarding
|
|
49
|
+
- `adapter status`, `diff`, `sync` — IDE adapter rule file synchronization
|
|
50
|
+
- Template recommendation heuristics with confidence scores
|
|
51
|
+
- Safety overwrites with automatic `.bak` backups
|
|
52
|
+
- `doctor --onboarding` diagnostics
|
|
53
|
+
|
|
54
|
+
### v2.7.0 — Website, Demo & Distribution System ✅
|
|
55
|
+
- Restructured homepage as a product conversion funnel
|
|
56
|
+
- Created 5 structured, copy-paste interactive demo workflow pages
|
|
57
|
+
- Documented comprehensive distribution and release workflows
|
|
58
|
+
- Added new SVG visual assets for onboarding and adapter sync flows
|
|
59
|
+
- Created docs-first examples for key developer workflows
|
|
60
|
+
- Updated sitemaps, model registries, and search indices
|
|
48
61
|
|
|
49
62
|
---
|
|
50
63
|
|
|
51
|
-
## 3.
|
|
64
|
+
## 3. Publishing Workflow
|
|
52
65
|
|
|
53
|
-
|
|
66
|
+
All releases follow this strict publishing checklist:
|
|
54
67
|
|
|
55
|
-
1.
|
|
56
|
-
2.
|
|
57
|
-
3.
|
|
58
|
-
|
|
59
|
-
|
|
60
|
-
```
|
|
61
|
-
4. [x] Build documentation static bundles cleanly:
|
|
68
|
+
1. Bump version in `package.json`
|
|
69
|
+
2. Run `npm run verify` (214+ assertions must pass)
|
|
70
|
+
3. Run `npm run docs:build` to verify documentation
|
|
71
|
+
4. Run `npm publish --dry-run` to review package hygiene
|
|
72
|
+
5. Set `MMDO_ALLOW_PUBLISH=true` and publish:
|
|
62
73
|
```bash
|
|
63
|
-
npm
|
|
64
|
-
```
|
|
65
|
-
5. [x] Perform a dry-run publish to review package hygiene:
|
|
66
|
-
```bash
|
|
67
|
-
npm publish --dry-run
|
|
68
|
-
```
|
|
69
|
-
6. [x] Set the required publication environment variable:
|
|
70
|
-
* **PowerShell:** `$env:MMDO_ALLOW_PUBLISH="true"`
|
|
71
|
-
* **Bash:** `export MMDO_ALLOW_PUBLISH=true`
|
|
72
|
-
7. [x] Publish the package to the public registry:
|
|
73
|
-
```bash
|
|
74
|
-
npm publish --access public
|
|
74
|
+
MMDO_ALLOW_PUBLISH=true npm publish --access public
|
|
75
75
|
```
|
|
76
76
|
|
|
77
77
|
---
|
|
78
78
|
|
|
79
|
-
## 4.
|
|
79
|
+
## 4. Upcoming: v2.8.0 — Interactive Dashboard & Plugin Hooks
|
|
80
|
+
|
|
81
|
+
* **Interactive TUI Status Dashboard**: Rich terminal UI for `status` and `workflow` commands
|
|
82
|
+
* **Plugin Hook System**: Pre/post hooks for `init`, `scan`, `memory`, and `workflow` commands
|
|
83
|
+
* **Custom Workflow Authoring**: User-defined workflow definitions beyond bundled registries
|
|
84
|
+
* **Adapter Auto-Detection**: Detect installed tools and automatically recommend adapter setup
|
|
85
|
+
|
|
86
|
+
---
|
|
87
|
+
|
|
88
|
+
## 5. Future: v3.0.0 — Unified Autonomous Co-Pilot Ecosystem
|
|
80
89
|
|
|
81
|
-
*
|
|
82
|
-
*
|
|
83
|
-
|
|
84
|
-
|
|
85
|
-
|
|
86
|
-
* **Configuration Upgrades**: Existing `.ai/` folders can be upgraded by running `init --force` to pull the new centralized registries (`.ai/models/` and `.ai/adapters/`) into the workspace.
|
|
90
|
+
* **Full Multi-Agent Orchestration**: Dynamic task handoffs between specialized agents
|
|
91
|
+
* **Distributed Registry Syncing**: Team-wide configuration synchronization
|
|
92
|
+
* **Cryptographic Proposal Signing**: Tamper-proof improvement proposals
|
|
93
|
+
* **Real-Time Collaboration**: Live workspace state sharing between agents and developers
|
|
94
|
+
* **Cloud-Native Intelligence**: Optional cloud-backed memory and feedback aggregation
|
|
87
95
|
|
|
88
96
|
---
|
|
89
97
|
|
|
90
|
-
##
|
|
98
|
+
## 6. Migration Notes
|
|
91
99
|
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
3. Explicit maintainer sign-off on local testing of the Android Expo template.
|
|
100
|
+
* **From any v2.x**: Run `npx multimodel-dev-os@latest init --force` to pull latest configuration files. Existing files are backed up automatically.
|
|
101
|
+
* **From v1.x**: See the [Migration Guide](/migration-guide) for the upgrade path.
|
|
102
|
+
* **Fresh install**: Simply run `npx multimodel-dev-os@latest init` — no prior setup required.
|