A production-grade agentic AI orchestration platform built with FastAPI, Ollama, and Qdrant. Agentica enables natural language interaction with real-world services — users can plan trips, check weather, search the web, manage schedules, and more through a unified AI agent layer.

Built as a professional portfolio project demonstrating real-world AI engineering, microservices architecture, and agentic AI orchestration skills.

A single user message like:

"Plan me a 3-day trip to Paris next weekend, find the cheapest flights from Tehran, book a hotel near the Eiffel Tower under $100/night, and tell me what to pack based on the weather forecast."

Gets automatically orchestrated across multiple specialized agents:

gateway-service (auth + routing) → agent-service (orchestrator) → travel-agent (flights + hotels) → weather-agent (Paris forecast) → llm-service (packing suggestions) → memory-service (save trip plan) → return complete plan to user

agentica/ ├── services/ │ ├── rag-service/ # Document ingestion, embedding, retrieval (port 8001) │ ├── llm-service/ # LLM inference, RAG-powered chat (port 8002) │ ├── memory-service/ # Conversation history, session management (port 8003) │ ├── agent-service/ # Agent orchestration, multi-step reasoning (port 8004) │ ├── gateway-service/ # Auth, routing, rate limiting (port 8000) │ ├── travel-agent/ # Flights, hotels, trains (port 8010) │ ├── weather-agent/ # Weather forecasts (port 8011) │ ├── search-agent/ # Web search, real-time information (port 8012) │ ├── finance-agent/ # Stocks, currency, crypto (port 8013) │ ├── news-agent/ # Latest news (port 8014) │ ├── calendar-agent/ # Schedule management, reminders (port 8015) │ ├── maps-agent/ # Directions, nearby places (port 8016) │ └── email-agent/ # Send/read emails (port 8017) ├── infra/ │ └── k8s/ # Kubernetes manifests (k3s) ├── docker-compose.yml └── README.md

• Docker + Docker Compose
• Python 3.11+

# clone the repository
git clone https://github.com/yourusername/agentica.git
cd agentica

# start all services
docker compose up --build

# pull the LLM model (first time only)
docker exec -it agentica-ollama ollama pull llama3.2:3b

cd services/rag-service

# create virtual environment
python -m venv venv
source venv/bin/activate

# install dependencies
pip install -r requirements.txt

# start qdrant
docker run -d -p 6333:6333 qdrant/qdrant

# run the service
uvicorn app.main:app --reload --port 8001

Each service exposes Swagger UI at /docs:

• Gateway: http://localhost:8000/docs
• RAG Service: http://localhost:8001/docs
• LLM Service: http://localhost:8002/docs
• Agent Service: http://localhost:8004/docs

Document → chunking (512 tokens, 50 overlap) → embedding (BAAI/bge-small-en-v1.5, 384 dimensions) → storage (Qdrant vector DB) Query → embed question → cosine similarity search (Qdrant) → retrieve top-k chunks → send question + chunks to LLM → return grounded answer

User message → gateway-service (auth + routing) → agent-service (task decomposition via LangGraph) → parallel tool calls to domain agents → results aggregated → llm-service generates final response → memory-service stores conversation → return to user

Internal services → HTTP/REST (current) → gRPC (planned) External clients → REST API via gateway-service Async tasks → RabbitMQ message queues Session state → Redis

# rag-service
cd services/rag-service
pip install pytest pytest-asyncio httpx
pytest tests/ -v

# llm-service
cd services/llm-service
pytest tests/ -v

• RAG Service
• LLM Service
• Docker Compose
• Memory Service
• Agent Service (LangGraph)
• Gateway Service

• Weather Agent
• Search Agent
• Travel Agent
• Finance Agent
• News Agent

• Auth Service (JWT + OAuth2)
• Kubernetes manifests (k3s)
• Prometheus + Grafana observability
• gRPC inter-service communication
• Calendar Agent
• Maps Agent
• Email Agent

Pull requests are welcome. For major changes please open an issue first.

MIT