OLLAMA_HOST=localhost:8000 ollama serve
mistral:latest
model on the host you are running Ollamaollama pull mistral:latest
docker-compose up
orca-mini
model: podman exec -it camel-assistant-ollama-1 ollama pull orca-mini
NOTE: this may take a while, as it needs to download about 2Gb of data from HuggingFace.
NOTE: URLs and hostnames can be configured in the application.properties
file or exported via environment variables. For instance
if using Qdrant in another host, you can set its host using the QDRANT_HOST
variable.
mvn clean package
podman run -d --rm --name qdrant -p 6334:6334 -p 6333:6333 qdrant/qdrant:v1.9.7-unprivileged
OLLAMA_HOST=localhost:8000 ollama serve
NOTE: make sure you have the mistral:latest
model available. If not, then download it using OLLAMA_HOST=localhost:8000 ollama pull mistral:latest
.
KAFKA_BROKERS=kafka-host:9092 java -jar ./assistant-ingestion-sink/target/quarkus-app/quarkus-run.jar
KAFKA_BROKERS=kafka-host:9092 java -jar ./assistant-ingestion-sources/plain-text-source/target/quarkus-app/quarkus-run.jar
KAFKA_BROKERS=kafka-host:9092 java -jar ./assistant-backend/target/quarkus-app/quarkus-run.jar
java -jar assistant-ui-vaadin/target/quarkus-app/quarkus-run.jar
To load PDF data (such as those from documentation, books, etc) into the QDrant DB, use the command:
cd assistant-cli && java -jar target/quarkus-app/quarkus-run.jar consume file /path/to/red_hat_build_of_apache_camel-4.0-tooling_guide-en-us.pdf
NOTE: you can download some PDFs from here.
You can load data from the Camel Dataset.
To download the dataset for data formats:
huggingface-cli download --repo-type dataset --local-dir camel-dataformats megacamelus/camel-dataformats
To download the dataset for components:
huggingface-cli download --repo-type dataset --local-dir camel-components megacamelus/camel-components
Use this command to load the dataset into the DB:
java -jar target/quarkus-app/quarkus-run.jar consume dataset --path ~/code/datasets/dataset/ --source org.apache.camel
Wait a few seconds after running the load command, and then check if the data is available in the Qdrant DB:
curl -X POST http://localhost:6333/collections/camel/points/scroll -H "Content-Type: application/json" -d "{\"limit\": 50 }" | jq .