I don't see any option of /clear in Codex I generally delete the chats and start new ones for it. I could not find any way of passing instruction with /compact. I do not use Hermes though so not sure. Thanks it was helpful
I used your /compact prompt and I got a very nice set of instructions that I used in a fresh new chat in Codex. Worked wonders even though I had to change the chats, so that's an extra step you have to take but the end result is the same.
Claude Chats will suffer from that issue because they have limited context and size limits for files. Th best way to work with it is to create Claude Code or Claude Cowork project and then point it to the kocal folder. That ways it can work through them with the context one by one.
Thank you so much @Caitlin McColl 🇨🇦 for sharing it
you’re welcome! :)
I had no idea that you can add instructions in /compact, that's something I'm going to use from now on. I wonder if it's the same in Codex and Hermes.
I don't see any option of /clear in Codex I generally delete the chats and start new ones for it. I could not find any way of passing instruction with /compact. I do not use Hermes though so not sure. Thanks it was helpful
I used your /compact prompt and I got a very nice set of instructions that I used in a fresh new chat in Codex. Worked wonders even though I had to change the chats, so that's an extra step you have to take but the end result is the same.
That is a great trick around it man, thanks for sharing it
When we are uploading documents, like pdfs, often there is a warning like below:
This chat has reached the 100-image limit (including PDF pages). Start a new chat to add more.
What are good practices to avoid hitting these limits?
Claude Chats will suffer from that issue because they have limited context and size limits for files. Th best way to work with it is to create Claude Code or Claude Cowork project and then point it to the kocal folder. That ways it can work through them with the context one by one.
For large self contained tasks I also like to ask Claude to have a subagent(separate context) handle it and report back.
Yes, that works... Generally Claude now does that mostly on its own.