Same agent Smaller bill. The tools and infrastructure to get more work from your AI. Fewer tokens. Less compute. Lower cost.
Testing and Developing Locally
There are two ways to run OpenLIT while you work on it:
- From source (recommended for code changes):
scripts/dev-realtime.shruns the UI with hot reload, the OTLP receiver, the realtime engine, and NATS on your machine, and starts ClickHouse in Docker if needed. See Run OpenLIT locally and the Local development docs. - Docker Compose: builds the OpenLIT image from this directory and runs it with ClickHouse and NATS. Use it to test the production image or the sample apps. The rest of this guide covers it.
The Compose setup is orchestrated by dev-docker-compose.yml and supports two modes:
| Mode | What starts | Use case |
|---|---|---|
| Default | ClickHouse + NATS + OpenLIT (UI, OTLP receiver, realtime engine) | Testing the OpenLIT image |
Full (--profile full) |
+ Controller + sample apps (OpenAI, Anthropic, Gemini, CrewAI, OpenAI Agents JS) | Testing the full end-to-end setup |
Running the Development Environment
Open a terminal and
cdinto thesrc/directory.UI only — build and start the OpenLIT dashboard:
docker compose -f dev-docker-compose.yml up --build -dFull setup — also build the controller and sample apps:
docker compose -f dev-docker-compose.yml --profile full up --build -dOnce the command completes, access the OpenLIT dashboard at http://127.0.0.1:3000 and sign in with
[email protected]/openlituser. Send OTLP to127.0.0.1:4318(HTTP) or127.0.0.1:4317(gRPC) with an API key from Settings → API Keys.
What's included in full mode
| Service | Description |
|---|---|
| ClickHouse | Time-series database for telemetry |
| NATS | JetStream bus for realtime signals (no published ports) |
| OpenLIT | Dashboard, first-party OTLP receiver, and realtime engine for threshold alerts (ports 3000, 4317, 4318) |
| Controller | eBPF-based LLM service discovery (built from openlit-controller/) |
| openai-app | Sample OpenAI chat app (from examples/) |
| anthropic-app | Sample Anthropic chat app (from examples/) |
| gemini-app | Sample Gemini chat app (from examples/) |
| crewai-app | Sample CrewAI agent app (from examples/) |
| openai-agents-js-app | Sample JavaScript/TypeScript OpenAI Agents SDK app with an in-process mock LLM (from examples/) |
Stopping the environment
# Stop and remove containers
docker compose -f dev-docker-compose.yml --profile full down
# Also remove stored data (ClickHouse + OpenLIT)
docker compose -f dev-docker-compose.yml --profile full down -v
Note: Include
--profile fullin the down command if you started with it, otherwise the profiled services won't be stopped.