How wolfsbane9513/influencer-ai works
Project Type & Purpose
• This is a backend-only, modular AI platform for automating influencer marketing campaigns.
• It exposes a FastAPI HTTP API, implements multi-agent orchestration (discovery, negotiation, contract) and integrates with external AI/voice services and a database.
Break Down the Core Architectural Layers
a. API Layer
• main.py / FastAPI app
• Routers: /api/webhooks, /api/monitor, plus “enhanced” variants
b. Orchestration Layer (Agents)
• orchestrator.py (master coordinator)
• discovery.py (creator matching)
• negotiation.py (voice-based deal closing)
• contracts.py (legal document generation)
• enhanced_* variants augment core agents
c. Service Layer
• database.py (persistence)
• embeddings.py (vector similarity)
• pricing.py (market data & rates)
• voice.py & enhanced_voice.py (ElevenLabs integration)
• conversation_monitor.py (real-time call tracking)
d. Data Layer
• Static files: data/creators.json, market_data.json, negotiation_tactis.txt
• Models: models/campaign.py (Pydantic schemas)
e. Config Layer
• config/settings.py (environment & feature flags)
f. External Integrations
• ElevenLabs API (voice calls)
• Groq API (AI strategy)
• Sentence Transformers (ML embeddings)
• PostgreSQL (optional, mocked by default)
Component Relationships & Data Flows
• Client → HTTP request → FastAPI router
• Router delegates to Orchestrator agent
• Orchestrator invokes:
– Discovery Agent → EmbeddingService → Sentence-Transformer → returns candidate list
– Negotiation Agent → VoiceService → ElevenLabs → live call transcripts
– Contracts Agent → (optionally PricingService → market_data) → generates contract document
• Each agent writes/reads state via DatabaseService
• Monitoring endpoints query DatabaseService or Agent state for real-time metrics
• Orchestrator collates results, returns final response or writes summary to DB
Architectural Patterns & Principles
• Layered Architecture: clear separation between API, Orchestrators (business logic), Services (infrastructure), Data
• Agent-Based Pipeline: each “phase” is encapsulated in its own agent class
• Asynchronous / Event-Driven: agents can run concurrently (asyncio.gather)
• Dependency Injection via config/settings
• Extensibility Points: enhanced_*.py modules, ability to add new agents or services