Mechanism for continuing an existing conversation about llm HOT 10 CLOSED

But how can it maintain track of a specific conversation?

There could be a command to start a new chat, then once started future chat commands maintain within that chat until you end the chat - or switch to another one.

Chats will make a lot more sense in the web interface, see:

• #4

Maybe there's a mode where a chat session is an interactive terminal process where you type in more text and hit enter to send a line at a time.

from llm.

The logging database schema should take chats into account as well.

from llm.

Demo of the freshly-merged #14 by @amjith:

$ llm 'What do otters like to eat?'
Otters are carnivores and they mainly eat fish, shellfish, crustaceans, and other aquatic animals. They have a varied diet, which includes clams, mussels, snails, crabs, crayfish, frogs, small mammals, and birds. They also enjoy eating insects and some vegetation occasionally.
$ llm -c 'where do they live?'
Usage: llm chatgpt [OPTIONS] [PROMPT]
Try 'llm chatgpt --help' for help.

Error: Invalid value for '-c' / '--continue': 'where do they live?' is not a valid integer.
$ llm 'where do they live?' -c
Otters are found in various parts of the world, inhabiting both freshwater and marine environments. There are 13 different species of otters, and their habitats vary depending on the species. 

River otters are commonly found in North America, from Alaska to Mexico, while sea otters live along the west coast of North America, from California to Alaska. Eurasian otters inhabit the rivers, lakes, and wetlands of Europe and Asia. 

Giant river otters, as their name suggests, are the largest of the otter species and can be found in the Amazon River basin of South America. 

Other species such as the smooth-coated otter, clawless otter and African clawless otter can be found in various parts of Asia and Africa, respectively.

It's a bit annoying that llm -c 'where do they live?' doesn't work, because the -c option needs to come at the very end. I wonder if we can improve that.

from llm.

Idea: llm -c/--continue could always act as a flag. That way you can use -c anywhere in the command.

Then llm --chat 44 could be an alternative option which requires a valid chat ID argument.

from llm.

I think --continue should reuse the same model as the logged conversation, unless the user specifies -m directly.

from llm.

I broke this - I forgot to set the default model so now 'llm "prompt"` complains that no model was specified.

from llm.

I'm having second thoughts about the --chat 123 option now - because maybe I want to use --chat as an option that starts an interactive chat UI instead?

from llm.

Options for an interactive chat mode:

• It's a new command - llm chat -m gpt-4 for example. This feels a bit odd since the current default command is actually llm chatgpt ... and llm chat feels confusing.
• It's part of the default command: llm --chat -4 starts one running.

Maybe the llm chatgpt command is mis-named, especially since it can be used to work with GPT-4.

from llm.

I named llm chatgpt that because I thought I'd have a separate command for bard and for llama and so-on, and because I thought the other OpenAI complete APIs (the non-chat ones, like GPT-3) may end up with a separate command.

from llm.

I'm closing this in favour of an issue to redesign the top-level set of commands.

from llm.