Federation

Universal Model Router for Governance-as-a-Service (GaaS)

"One API for 50+ model providers. Quality-first routing with enterprise governance."

Features

Universal Router — Route tasks to optimal providers across 50+ options
Quality-First — Weights: Quality (50%) > Speed (30%) > Cost (10%) > Reliability (10%)
Enterprise Governance — SOC2, GDPR, HIPAA compliance built-in
Cost Optimization — Automatic savings vs single-provider strategies
Health Monitoring — Circuit breakers, failover, real-time dashboards
Multi-Tenant — SaaS-ready with tenant isolation

Quick Start

Installation

# Clone the repository
git clone https://github.com/hummbl-dev/hummbl-agent-federation.git
cd hummbl-agent-federation

# Install with Poetry
poetry install

# Or with pip
pip install -e .

Initialize a Project

# Create a new federation project
federation init

# This creates:
# config/
# └── providers.yaml    # Your provider configurations

Route Your First Task

# Route a task (dry-run)
federation route "Implement JWT authentication" --dry-run

# Output:
# ┌─ Routing Decision ──────────────────────────────┐
# │ deepseek                                        │
# │                                                 │
# │ Confidence: 92%                                 │
# │ Est. Cost: $0.05                                │
# │ Est. Latency: 2100ms                            │
# └─────────────────────────────────────────────────┘

# See why this provider was chosen
federation route explain "Implement JWT authentication"

List Providers

federation provider list

# Output:
# ╭─ Federation Providers ────────────────────────╮
# │ Provider    Status  Models  Latency  Cost/Mtok │
# ├────────────────────────────────────────────────┤
# │ 🅾️ OpenAI    ●       12      1.2s    $2.50     │
# │ 🅰️ Anthropic  ●        8      1.5s    $3.00     │
# │ 🐋 DeepSeek   ●        4      2.1s    $0.14     │
# │ ⚡ Groq       ●       15      0.3s    $0.59     │
# │ 🏠 Ollama     ●        3      0.1s    $0.00     │
# ╰────────────────────────────────────────────────╯

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         TASK INPUT                               │
│    "Implement authentication middleware with JWT"               │
└────────────────────┬────────────────────────────────────────────┘
                     │
                     ▼
┌─────────────────────────────────────────────────────────────────┐
│                     FEDERATION ROUTER                            │
│                                                                  │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────────────┐ │
│  │   Intent    │  │   Filter    │  │        Score            │ │
│  │  Classifier │─▶│   Engine    │─▶│   Quality: 50%          │ │
│  │             │  │             │  │   Speed:   30%          │ │
│  │ code_impl   │  │ GDPR? SOC2? │  │   Cost:    10%          │ │
│  └─────────────┘  └─────────────┘  │   Reliability: 10%      │ │
│                                     └─────────────────────────┘ │
└────────────────────┬────────────────────────────────────────────┘
                     │
        ┌────────────┼────────────┐
        │            │            │
        ▼            ▼            ▼
┌──────────────┐ ┌──────────┐ ┌─────────────┐
│   DeepSeek   │ │   Groq   │ │   OpenAI    │
│  $0.05/task  │ │ $0.12/t  │ │  $2.50/task │
│  Quality:88% │ │ Quality  │ │  Quality:95%│
└──────────────┘ └──────────┘ └─────────────┘

Provider Support

Tier 1: Frontier Labs

OpenAI — GPT-4o, o1, o3-mini
Anthropic — Claude 3.5/4 Sonnet, Opus
Google DeepMind — Gemini 1.5/2.0 Pro
Meta — Llama 3.1/3.2/3.3 (via aggregators)
xAI — Grok-2, Grok-3
Cohere — Command R, Embed
Mistral AI — Mistral Large, Codestral

Tier 2: Chinese Frontier

DeepSeek — V3, R1 (reasoning)
Alibaba — Qwen2.5 series
Moonshot AI — Kimi (long context)
And 7 more...

Tier 3: Cloud Aggregators

OpenRouter — 300+ models, one API
Together AI — Fast inference
Groq — 800+ tokens/sec
Fireworks — Production inference
And 6 more...

Tier 4-7: Cloud Marketplaces, Specialized, Open Source, Emerging

50+ total providers supported.

Governance-as-a-Service

Compliance Policies

policy:
  name: soc2-strict
  rules:
    - all_requests_logged: true
    - pii_exclusion_required: true
    - data_residency: [us, eu]
    - retention_days: 2555  # 7 years
    - encryption: aes-256-gcm

Enterprise Features

Immutable Audit Trails — Every request logged with provenance
Data Residency — Route to providers in specific regions
PII Detection — Automatic exclusion of sensitive data
RBAC — Role-based access control per tenant
Budget Controls — Hard limits with alerts

Development

Setup

# Install dev dependencies
poetry install --with dev

# Run tests
poetry run pytest

# Run with coverage
poetry run pytest --cov=src

# Format code
poetry run black src/ tests/
poetry run ruff check src/ tests/

# Type checking
poetry run mypy src/

Project Structure

federation/
├── src/
│   ├── registry/          # Provider registry
│   │   ├── models.py      # Provider dataclasses
│   │   ├── loader.py      # Config loading
│   │   └── store.py       # SQLite/Redis storage
│   ├── engine/            # Routing engine
│   │   ├── classifier.py  # Intent classification
│   │   ├── router.py      # Main routing logic
│   │   └── cost.py        # Cost tracking
│   ├── adapters/          # Provider API adapters
│   ├── governance/        # GaaS policy engine
│   ├── telemetry/         # Observability
│   ├── security/          # Vault, encryption
│   ├── tenant/            # Multi-tenancy
│   ├── cli/               # Command line interface
│   └── patterns/          # MapReduce, Debate, etc.
├── tests/                 # Test suite
├── docs/                  # Documentation
└── config/                # Provider configurations

Roadmap

Phase 1: Foundation ✅

Provider registry
Routing engine
CLI scaffolding
Provider adapters (in progress)

Phase 2: Intelligence

Adaptive learning
Real-time dashboard
Cost optimization
A/B testing framework

Phase 3: Scale

Kubernetes deployment
Multi-region routing
Advanced security
Workflow patterns

Phase 4: Ecosystem

SDKs (Python, TS, Go)
IDE extensions
Consulting methodology
Training curriculum

Business Model

Service Tiers

Tier	Price	Target	Features
Community	Free	Developers	1K requests/mo, 5 providers
Pro	$99/mo	Startups	100K requests, 20 providers
Enterprise	Custom	Large orgs	Unlimited, SOC2, consulting
Consulting	Project	Transformation	Base120 implementation

Revenue Split

Platform Usage: 70%
Consulting Services: 20%
Education: 10%

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT License — See LICENSE

Contact

GitHub: hummbl-dev/hummbl-agent-federation
Email: reuben@hummbl.io

<p align="center"> <em>Built with ❤️ by the HUMMBL Agent Federation</em> </p>

Name	in19-harm-minimization-via-negativa
Description	Apply IN19 Harm Minimization (Via Negativa) to improve by removing harmful elements rather than adding beneficial ones.

in19-harm-minimization-via-negativa

SKILL.md