Camel Assistant

Trying it using Docker Compose

1. Start an OpenAI compatible API in a host accessible from the containers

OLLAMA_HOST=localhost:8000 ollama serve

2. Pull the mistral:latest model on the host you are running Ollama

ollama pull mistral:latest

3. Start the containers using Docker Compose

docker-compose up

4. Wait for everything to be up and then pull the 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.

5. Proceed to the Loading Data section for details about how to load data

Trying it manually

Requirements

• A Kafka instance up and running and able to receive remote requests.
• Podman installed and running (locally or remote)
• Ollama installed and running (locally or remote)

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.

Steps

1. Build the project

mvn clean package

2. Launch Qdrant:

podman run -d --rm --name qdrant -p 6334:6334 -p 6333:6333 -v $(pwd)/qdrant_storage:/qdrant/storage:z qdrant/qdrant:v1.9.7-unprivileged

3. Launch Ollama:

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.

4. Launch the ingestion sink:

KAFKA_BROKERS=kafka-host:9092 java -jar ./assistant-ingestion-sink/target/quarkus-app/quarkus-run.jar

5. Launch the ingestion source:

KAFKA_BROKERS=kafka-host:9092 java -jar ./assistant-ingestion-sources/plain-text-source/target/quarkus-app/quarkus-run.jar

6. Launch the backend:

KAFKA_BROKERS=kafka-host:9092 java -jar ./assistant-backend/target/quarkus-app/quarkus-run.jar

7. Launch the UI:

java -jar assistant-ui-vaadin/target/quarkus-app/quarkus-run.jar

Loading Data

Loading PDFs

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.

Loading Datasets

You can load data from the Camel Dataset.

To download the dataset for data formats, components, EIPs and languages:

huggingface-cli download --repo-type dataset --local-dir camel-dataformats megacamelus/camel-dataformats
huggingface-cli download --repo-type dataset --local-dir camel-components megacamelus/camel-components
huggingface-cli download --repo-type dataset --local-dir camel-eips megacamelus/camel-eips
huggingface-cli download --repo-type dataset --local-dir camel-languages megacamelus/camel-languages

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/camel-dataformats/ --source org.apache.camel
java -jar target/quarkus-app/quarkus-run.jar consume dataset --path ~/code/datasets/camel-components/ --source org.apache.camel
java -jar target/quarkus-app/quarkus-run.jar consume dataset --path ~/code/datasets/camel-eips/ --source org.apache.camel
java -jar target/quarkus-app/quarkus-run.jar consume dataset --path ~/code/datasets/camel-languages/ --source org.apache.camel

Checking if the data was loaded

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 .

Building

To build the containers locally:

mvn clean package -Dquarkus.container-image.build=true