[Experimental] Memory-enabled agent

[enyst]

End-user friendly description of the problem this fixes or functionality that this introduces

• Include this change in the Release Notes. If checked, you must provide an end-user friendly description for your change below

---

Give a summary of what the PR does, explaining any non-trivial design decisions

This PR proposes a memory-enabled agent. It integrates the approach implemented by @khushvind and aims to implement a (bit larger) part of a memory system than we have previously:

• condense messages when necessary / with an experimental option for LLM's choice
• recall memories
• use some form of long-term memory - embeddings for recall (llama-index and/or litellm, both available for testing)

Experiment:

• inspired by the Generative Agents paper and MemGPT paper, it aims to let the agent manage its own memory. Not just us deciding when to re-route stuff to condense, rather have it aware of its memory system and let it decide what is important enough to keep, when and where.

Rough schema of cognitive architecture illustrated among others in Cognitive Architectures for Language Agents:

---

Link of any specific issues this addresses