- Add Mem0MemoryService to enable conversation memory persistence
- Add get_initial_greeting function to create personalized greetings
- Update pipeline to include memory service between user context and LLM
- Add support for both cloud-based (Mem0 API) and local configurations
- Update system instructions to include memory-related guidance
- Modify on_client_connected handler to fetch and use personalized greetings
- Update documentation with Mem0 setup and usage instructions
Make `LLMUserAggregator` push the `LLMSetToolsFrame`s, in case a speech-to-speech service that needs to handle the frame itself—like `OpenAIRealtimeLLMService`—is downstream. As far as I can tell, pushing `LLMSetToolsFrame` should otherwise have no unwanted side effects.
Push `TranscriptionFrame`s upstream, to be handled by the user context aggregator. This will require at least a couple of other changes:
- Updating examples to put transcript processors upstream from `OpenAIRealtimeLLMService`
- Maybe figuring out a way to preserve backward compatibility with existing pipelines that put transcript processors downstream from `OpenAIRealtimeLLMService`
- Updating `OpenAIRealtimeLLMService` to ignore new received context frames, since the upstream user context aggregator will generate those after each newly-added user message; hopefully nobody was reliant on the old behavior of resetting the conversation upon receiving a new context!
The reason for its `system_instruction` argument was to support usage with LLMs where you might pass the system instruction as a parameter to the `LLMService` rather than specifying it in the context.
But as I thought about it more I became unconvinced that the `system_instruction` argument was really beneficial:
- If you specified your system instruction in your context in the first place, it'll still be there when you read messages for persistent storage
- If you didn't specify your system instruction in the context and instead passed it in as an `LLMService` parameter, you most likely *don't* want it to be in the context when you read messages for persistent storage
- ...and if you really really do need to inject it at the start of the context, it's quite easy to do anyway
And if we remove the `system_instruction` argument from `get_messages_for_persistent_storage()`, then it's essentially just `get_messages()`.
