Autonomous omnichannel marketing orchestration platform with LLM agents, real-time budget reallocation, counterfactual what-if simulation, policy-as-code approvals, and a natural-language command surface for marketing teams running Google Ads, Meta, TikTok, Snap, LinkedIn, Amazon DSP, WhatsApp, email, SMS, and cold-call programs.
This repo is designed to be highly discoverable for searches around:
- autonomous marketing orchestrator
- agentic marketing budget optimization
- LLM marketing operations platform
- real-time media budget reallocation
- omnichannel campaign orchestration
- marketing mix optimization with AI agents
- natural-language ad operations dashboard
Modern marketing teams still operate like disconnected desks:
- paid social moves budget too late
- CRM and lifecycle teams cannot see media pressure in real time
- call, SMS, and WhatsApp sequences are detached from media pacing
- nobody can answer a live question like:
What happens if I move $10K from Meta to TikTok at 2 PM?
marketing-orchestrator-llm turns that into one operator-grade control plane.
Each major channel gets an evaluation agent that watches:
- ROAS
- CPA
- CTR
- CPM
- frequency
- demand temperature
- data reliability
- saturation risk
Those agents submit budget proposals to a coordinator layer.
The Python orchestration core uses a solver-backed optimization path to:
- rebalance a fixed budget pool
- protect channel caps
- react to short-horizon marginal returns
- preserve policy safety rails
Operators can issue commands like:
increase Snap budget by 15% for the next 4 hours because we have a flash sale
The platform parses the command, extracts channel intent, and validates the move against policy constraints.
The counterfactual engine simulates reallocation decisions and reports estimated revenue delta before execution.
The repo includes versioned funnel logic and workflow assets for:
- SMS
- cold-call escalation
- n8n workflow handoffs
- Python / FastAPI for orchestration APIs and optimization
- TypeScript / Next.js for the operator dashboard
- SQL for attribution and KPI rollups
- Rego for policy-as-code
- JSON for n8n workflow definitions
- Bash for deployment automation
- Go for Kafka clickstream ingestion
- Rust for Meta CAPI batching
- Java for Flink stream jobs
- C# for Azure / Microsoft Ads ingestion
marketing-orchestrator-llm/
├── apps/dashboard/
├── services/orchestrator_api/
├── services/go-kafka-consumer/
├── services/meta-capi-rust/
├── services/flink-kpi-java/
├── services/msads-ingest-csharp/
├── policies/
├── sql/
├── n8n/workflows/
├── docs/
└── tests/
pip install -r services/orchestrator_api/requirements.txt
uvicorn services.orchestrator_api.app.main:app --reload --port 8000cd apps/dashboard
npm install
npm run devOpen:
- API:
http://localhost:8000 - Dashboard:
http://localhost:3000
GET /healthGET /api/statePOST /api/reallocatePOST /api/commandPOST /api/what-ifGET /api/funnel
- sees a live budget efficiency score
- reviews incremental ROI without channel silos
- issues Black Friday or flash-sale instructions in plain English
- audits every budget decision through a replay token
- stops manually moving spend across 10+ surfaces
- sees which channels should be increased or throttled now
- gets policy-aware reallocation logic
- keeps SQL, counterfactual framing, and replayability in the same platform
This repo was verified locally with:
pip install -r services/orchestrator_api/requirements.txt
python -m pytest tests -q
cd apps/dashboard
npm install
npm run buildMore detail is in docs/final-audit.md.
Built and maintained by Sameer Alam.