Long messages cause an error

[lincheoll]

Describe the bug

Long messages cause an error.

To Reproduce

Using PostgreSQL as the database, sending a message longer than 255 characters causes an error.

Alternatively, if the most recent message is too long (an error occurs in factsProvider), the issue arises due to the message length. Since it uses the levenshtein function, the error message:

"levenshtein argument exceeds maximum length of 255 characters"

appears when a long message is processed. This happens immediately when sending a long message at once. Applying slicing resolves the issue, but I'm not sure if that's the right approach.

It would be great if someone could fix this.

I'm curious about the best way to handle this, though I haven't tested it on other databases.

Expected behavior

Long texts or past chat history should not cause errors.

Screenshots

Additional context

I don’t think this happened before… I’m not sure. I don’t know where to look.

[alois8]

+1
(I have this in v0.25.6-alpha.1)

[lincheoll]

@alois8
After checking, such an error does not occur in v0.1.9.
However, there seem to be other issues…

[odilitime]

Encountering this is a v0.1.9 copy

[tercel]

modify to below can work well.
levenshtein( left($1, 255), left(content_text, 255) )

[odilitime]

can confirm this worked

[salacoste]

still a problem at #v0.25.6-alpha.1 when switched on postgres

[salacoste]

modify to below can work well. levenshtein( left($1, 255), left(content_text, 255) )

at which file, sorry?

[nabeelraza-7]

modify to below can work well. levenshtein( left($1, 255), left(content_text, 255) )

at which file, sorry?

You will have to change getCachedEmbeddings function in your database adapter. In my case, it was adapter-postgres/src/index.ts and that worked.

[kingcozz]

how does this work if it cuts off at 255? dont we need the entire thing?

[odilitime]

it's just index used for searching, messages with more than 255 character won't have all their keywords indexed

[daniel-trevino]

I can also confirm @tercel 's fix is working on my fork. I can open a PR with the fix if nobody is up for it.

What's the best approach to patch the bug? On the new community plugins or directly in main? cc @odilitime

[odilitime]

Adapter changes happens in elizaos-plugins, core changes here in eliza

[retdude]

Has this been fixed yet? I have the same issue on my end as well. I opened an issue over here that describes my issue. It happens always, the moment I run pnpm start. Not even just when sending a long message:

elizaos-plugins/adapter-postgres#1

[xircusfounder]

any status on this fix?

[AlejoLovallo]

I have added the same fix in the adapter repository, they forgot to added it when they make the reorganization

• Circuit Breaker error elizaos-plugins/adapter-postgres#1
  Please merge it! Fix long messages error elizaos-plugins/adapter-postgres#4

[lincheoll]

In v2, the plugin has been separated into a different codebase, so I will close this issue.