-
Notifications
You must be signed in to change notification settings - Fork 1.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
JSON parser errors when retrieving C++ code from archival memory. #729
Comments
The above shows only some of the C++ code from archival memory. This is from a proprietary code base so I am not at liberty to freely share more code. |
Just confirmed via pytest it looks like our current JSON parser breaks on that string, will add a patch for it now:
|
Patch for the specific LLM output in your trace is here (so you shouldn't see that exact error anymore): #739 |
I do not know if this has been reported yet, but when I talk to the bot in English, there are no JSON errors. However, as soon as I speak in French (thus with accents and characters that do not exist in English), I get a lot of errors. There may be an issue due to this. Surprising in 2024... |
This issue has been automatically closed due to 60 days of inactivity. |
Describe the bug
I have loaded some C++ source code using
memgpt load directory ...
. So far I am loading just one component directory that has just a few source files so that I can experiment and work through issues. The first issue is that the JSON parser has difficulty parsing the response after searching archival memory. My hypothesis is that the code contains character sequences that are not properly encoded/escaped which causes the stringified JSON to be malformed.I will attach a larger capture from the session, but the initial part of the error message is:
Please describe your setup
memgpt version
? (eg "0.2.4")I am using
pymemgpt-nightly
from the night of Dec 27th/28th.memgpt version
doesn't reflect this -- perhaps you can add some suffix to the semver string?How did you install memgpt?
Basically I did:
pip install pymemgpt-nightly
However this was after a
pip uninstall pymemgpt
and before that I had donepip install pymemgpt
followed bypip install pymemgpt[local]
.Describe your setup
What's your OS (Windows/MacOS/Linux)?
Apple Powerbook M2 64Gb ram running Sonoma 14.2.
How are you running
memgpt
? (cmd.exe
/Powershell/Anaconda Shell/Terminal)Terminal session inside VS Code.
Screenshots
If applicable, add screenshots to help explain your problem.
Additional context
I created the data source using this command:
I then ran
memgpt
and did a/attach
and selectedtest_action_state_machine
.If you're not using OpenAI, please provide additional information on your local LLM setup:
Local LLM details
Everything running locally:
The text was updated successfully, but these errors were encountered: