Integration with open interpreter

[ian-reay]

I've been using open interpreter ( https://github.com/KillianLucas/open-interpreter ) quite a bit. I have an engineering/development background so I've been using it to

• Troubleshoot system stability problems - often scanning system logs
• Analyze application log files that are sometimes gigs in size
• Analyze data in spreadsheets

I've also been talking to people in non-engineering areas such as:

• Legal: Analyzing contracts stored on a network share. Conversing with my documents that sometimes are pretty large.
• Sales & Markting: Analyzing market data
• Services: analyzing and creating statements of work and and project plans.

In these scenarios, I can ask open interpreter to analyze data and files. But due to the restricted context windows, open-interpreter currently leverages python scripts very heavily where it will use python to grep, search for keywords, and other scenarios that lack the capabilities of natural language search and a viable long term memory of conversations.

Being able to load a documents ( log files, crash dumps, spreadsheets, contracts, market analysis, project plans ) into your memory and then being able to interpret it effectively holds huge promise. Right now I can get exceptional results when I can fit my problem into gtp 4's 32k context window. But in the above cases that 32k tokens run out very quickly.

[ChuxiJ]

I am working for it

see： timedomain-tech/open-creator#15

[ChuxiJ]

Aha~ mini MemGPT It works

[TheMavel]

Have you found a way to use Mem-gpt with open-interpreter? I would love this.
I was thinking about setting up a Mem-gpt mistral 7b model locally as an api for the code-interpreter but I could not get it to work.