How to Deploy Cognita with Docker Compose: Step-by-Step Tutorial

Deploying Cognita: step-by-step

In this tutorial you will stand up a complete Cognita RAG stack with Docker Compose — including Qdrant for vector storage and Ollama for fully local LLM inference — then ingest your first documents and chat with them through the web UI.

Step 1 — Clone and configure

Start by cloning the repository and copying the environment template:

git clone https://github.com/truefoundry/cognita.git
cd cognita
cp compose.env .env

Open .env and set your POSTGRES_USER, POSTGRES_PASSWORD, and pick an OLLAMA_MODEL (for example llama3.2:3b). The compose file wires up PostgreSQL for metadata, Qdrant for vectors, Ollama for embeddings and generation, and the Infinity server for embedding inference.

Step 2 — Start the stack

Launch the database and vector services first, then enable the optional Ollama profile:

docker compose --profile ollama up -d

Wait for the health checks to pass — PostgreSQL and Qdrant must be healthy before the backend starts. The first Ollama run pulls the model you selected, which can take a few minutes depending on your connection.

Step 3 — Create a collection and add data

Open the Cognita UI at http://localhost:8080. Create a new collection, then add a datasource. You can point it at a local PDF, a website URL, or any of the 30+ supported connectors:

Trigger ingestion and watch the progress in the UI. When it completes, the chunks are embedded and stored in Qdrant:

Step 4 — Ask questions and evaluate

Head to the chat interface, pick your collection, and start asking. Cognita runs hybrid search with reranking before passing context to the LLM:

Use the evaluation panel to compare retrieval quality across different embedding and reranker configurations — this is where production-ready answers come from.

Hardware notes

That is it — a fully local, production-oriented RAG stack in four steps. Cognita is a great pick when you want full control over every stage of retrieval augmented generation without building the plumbing yourself.