Any tips on handling conversation history for chatbots?

[bomb]

Experimenting with llama-cpp-python and 7b gguf models like llama2 and mistral. Smooth sailing, but hit a snag with history-dependent queries. The generator doesn't track history, so input grows fast, affecting processing time and response quality. I tried summarizing with another generator, but messy results. Any standard tips for chatbot history?

 

[Catalina]

Experimenting with llama-cpp-python and 7b gguf models like llama2 and mistral. Smooth sailing, but hit a snag with history-dependent queries. The generator doesn't track history, so input grows fast, affecting processing time and response quality. I tried summarizing with another generator, but messy results. Any standard tips for chatbot history?

Try chunking the input, summarizing with ROUGE/BLEU, or using history-aware models/beam search. Train on dialogue data, set a max history length, and monitor memory usage.

 

[Kali]

I delete my conversation history. I do not know if the conversation is completely deleted or not, however, I do not like to see the last communication on my dashboards. Every time I start a chat, it is always a new chat. I do this for all chatbots I use.

 

[Killacountai43]

I delete my conversation history. I do not know if the conversation is completely deleted or not, however, I do not like to see the last communication on my dashboards. Every time I start a chat, it is always a new chat. I do this for all chatbots I use.

I do the same thing, as I don't see the point of having the information lingering around in Chatgpt. Also, I don't want anyone going through my search history should they somehow get hold of my Chatgpt account.