Refer to package README for most up-to-date instructions.
git clone [email protected]:robusta-dev/holmesgpt.git
cd holmesgpt
poetry install
Refer to HolmesGPT documentation for datasource and LLM configuration options.
Below is an example using Prometheus and Claude Sonnet 4 on AWS Bedrock:
HOLMES_HOST=0.0.0.0
HOLMES_PORT=5050
LOG_LEVEL=INFO
AWS_PROFILE=default
AWS_REGION=us-east-1
LITELLM_MODIFY_PARAMS=true
MODEL=bedrock/us.anthropic.claude-sonnet-4-20250514-v1:0
PROMETHEUS_URL=http://localhost:9090
OPENSEARCH_URL=http://localhost:9200
INDEX_PATTERN=logs-otel-*
HolmesGPT added experimental support for AG-UI in this PR. Launch the server on your localhost:
poetry run python experimental/ag-ui/server.py
HolmesGPT AG-UI chat server should now be available at http://localhost:5050/api/agui/chat.
Below is an example for OSD v3.3. Refer to the DEVELOPER_GUIDE for most up-to-date instructions.
git clone [email protected]:opensearch-project/OpenSearch-Dashboards.git
cd OpenSearch-Dashboards
yarn osd bootstrap
In your opensearch_dashboards.yml, enable AI Chat features and configure chat.agUIUrl to use the HolmesGPT chat enpoint:
# @experimental Set the enabled value to true and provide an AG-UI agent to enable chat
chat:
enabled: true
chat.agUiUrl: 'http://localhost:5050/api/agui/chat'
# @experimental Set the value to true to enable context provider
contextProvider:
enabled: true
# Enable the new explore experience (required for AI features)
explore.enabled: true
yarn start
This fork of the opentelemetry-demo contains a working example of HolmesGPT, OSD, and Prometheus. It includes synthetic metrics, logs and traces.
- Modify the .env file at the repository root with your HolmesGPT environment variables as per above installation instructions.
- Modify src/dashboards/config/opensearch-dashboards.yml as per above installation instructions.
git clone [email protected]:kylehounslow/opentelemetry-demo.git
cd opentelemetry-demo
docker compose up -d