epic: conversation pipeline

[mhordynski]

Feature description

Ragbits should provide streamlined way to create chat / conversation pipelines.

It should handle:

• Input / Output models for pipeline
• Allow for metadata / external method to determine current user
• Allow for plugin for conversation history persistence (feat: add history persistence component #255)
• Allow for calling tools like DocumentSearch (either by function calling or linear) or HistoryCompressor
• If any of the tools, in this example DocumentSearch, returns any additional context such as DocumentSearch' Element it should be part of output schema from the pipeline.
• Text output from the model should be streamed

Motivation

Ragbits is commonly used to generate chatbot experiences, thus this process should be streamlined and provide nice integration with ecosystem.

Additional context

Implementation should start with #316

Draft of Input schema:

• Prompt Input
• Message of the user, history

Pipeline should accept main Prompt instance that is used in conversation with LLM. Output of the pipeline is inferred from this prompt.