Skip to content

SIF & AI Agents Overview

The Synthetic Intelligence Framework (SIF) is ALwrity's agent orchestration system — a team of specialised AI agents that collaborate to plan, execute, and audit your daily content marketing operations.

The Agent Ecosystem

ALwrity runs 8 specialised agents grouped into two tiers:

Committee Agents (6)

These agents are polled every day to propose tasks for the daily workflow. They are the engine that generates your actionable task list.

Agent Registry Key Role
Content Strategy Agent content_strategist Identifies content opportunities and strategy gaps
Strategy Architect Agent strategy_architect Suggests strategic pivots from SIF semantic analysis
SEO Optimization Agent seo_specialist Flags SEO issues and optimisation opportunities
Social Amplification Agent social_media_manager Recommends social engagement and distribution actions
Competitor Response Agent competitor_analyst Alerts on competitor content moves and gaps
Content Gap Radar Agent content_gap_radar Scores content opportunities by combining SERP data, competitor deep-dives, and trend momentum

Independent Agents (2)

These agents are not polled for daily tasks — they operate alongside the committee to monitor, audit, and provide specialised signals.

Agent Registry Key Role
Trend Surfer Agent trend_surfer Identifies emerging market trends and suggests timely content angles
Content Guardian Agent content_guardian Watchdog that audits committee proposals, evaluates agent behaviour, and flags coverage gaps

Orchestrator

The StrategyOrchestratorAgent is the master coordinator. It receives the agents' proposals, runs deduplication, enforces pillar coverage, and delegates strategic execution tasks back to sub-agents when needed.

How Agents Are Initialised

Agents are created per-user by the ALwrityAgentOrchestrator when a user completes onboarding. Each agent can be individually enabled or disabled via AgentProfile settings:

profiles = db.query(AgentProfile).filter(AgentProfile.user_id == user_id).all()
enabled = {p.agent_key: bool(p.enabled) for p in profiles}

If a profile is missing or enabled is None, the agent defaults to active.

Data Flow

Onboarding Complete
        │
        ▼
Orchestrator Initialised
    ┌───┴───┐
    │       │
    ▼       ▼
Committee    Independent
Agents (6)   Agents (2)
    │           │
    │           ├─ Trend Surfer → trend_signals event
    │           └─ Content Guardian → quality_audit event
    │
    ▼
Daily Workflow
  ┌── Proposal Phase (parallel)
  ├── Deduplication
  ├── Pillar Coverage
  ├── LLM Fallback
  ├── Contextuality Validation
  └── SIF Indexing

Agent Activity Feed

Every agent action — proposals, audits, errors, initialisations — is logged to the AgentActivity event feed. This powers the real-time Team Activity dashboard and enables persistent alert tracking.

Key Design Principles

Principle Description
Separation of concerns Committee agents propose; independents audit. ContentGuardianAgent never proposes daily tasks alongside the committee.
Deterministic scoring Audit scoring uses heuristics (keyword presence, length, ratios), not LLM calls — fast, deterministic, and predictable.
Parallel execution Committee proposals run concurrently via asyncio.gather.
Per-user enable/disable Each agent can be toggled independently through AgentProfile settings.
No cross-agent prompt pollution Agents do not receive each other's outputs in their prompts. The ContentGuardianAgent's audit is stored separately and never fed back into committee prompts.