Provide a few default fallback methods to call OpenAI API or run some Langchain pipeline to query a file

[mgv99]

[jcabot]

For instance, we could have at least:

• a simple one where we just directly ask chatgpt the exact same question (as part of the configuration the bot creator should addd an openai key)
• another one where we let the bot creator to upload (or point) to a pdf and have a langchain script that uses that pdf to try to answer the quesion

[mgv99]

Done with LLM and RAG (#77)