fix(k8s): chroma initcontainer

[cmdoret]

Objective: k8s deployment working

Proposed changes

This PR addresses several bugs that broke the k8s deployment:

• outdated langchain parameters
• wrong variable in default answer_template
• incorrect kustomize overlay fieldpaths
• outdated default openai model (davinci)
• wrong service type for chroma
• incorrecly specify protocol in hostname env vars
• chunk insert query in insert_triples to prevent crashing graphdb service on upload

Limitations

This PR only gets the manifests working, but there is much to improve on. The different shortcomings are tracked as issues:

• limited configuration options through manifests: https://github.com/sdsc-ordes/kg-llm-interface/issues/33
• no single deployment-level configuration file: https://github.com/sdsc-ordes/kg-llm-interface/issues/31
• no secret management (openai key, graphdb/chroma credentials): https://github.com/sdsc-ordes/kg-llm-interface/issues/30

How to try it:

• change the openai key in k8s/base/kg-llm/params.env to your own
• deploy all services with kubectl apply -k k8s/overlays/custom-config
• after all init containers have finished running, open a port forawrd to the kg-llm service and query it with localhost:<port>/ask?question="a question about pokemons"