-
Notifications
You must be signed in to change notification settings - Fork 0
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
seed_type memorize
and recall
#23
Comments
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
This is just a low-efficiency in-memory representation. ProfileFilesystem should implement a higher-performance version. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
Just lightly tested currently. Doesn't do much until `recall` is wired up. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
If there were fewer than k items then it would return none of them. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 2, 2023
It recalls memories previously stored with `memorize`. Currently `memorize` doesn't persist because ProfileFilesystem just holds the memories in memory. But once that is implemented it will actually be useful... Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…in a local db. Currently it has a number of limitations including not actually persisting the text associated with each embedding, and a number of other TODOs in the code. But it does roughly work. Part of #23
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…f recall and memorize if mocking. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…BY_MODEL. Currently it's unncessary, but it will be soon when we start persisiting to disk the text. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
It's up to the callsite to set defaults. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…bedding. For now it's just a JSON file that's written out to filesystem fully on each update. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…rom filesystem. We were initializing the _metadata to {} in constructor so it was never empty. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
If so, double the size of the index before inserting. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…bedding can be used. Where an embedding is found, it will use embedding.text. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
Embeddings are so long that printing them all out in weird json was unncessary and overwhelming. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
…default_memory` to make it more obvious what context they're in. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
Turns out it doesn't require complex new machinery; `input.default` was also handled similarly. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
If provided, then it will be used instead of the `memory` env variable. Part of #23.
jkomoros
added a commit
that referenced
this issue
Jul 3, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
A local associative memory plugged into the library and matching a typespec.
value
for memorize is an array of text, process them in parallel (combining into a single request) instead of sequentially (which adds a network round trip time * items.length)_default
) to something more dsitinctivee that it's being used in a memory context, not a proflie context (similar to ids havingc_whatever
testing recall seed
by actually using real (cached) embeddings for those values. This is somewaht important to verify the sorting is actually correct and not backwards....profile/memory/MEMORY-NAME/${normalized_embedding_model_name}/hsnw.db
. This requires the embedding_model_name to not have any illegal path charactersrecall.k
seed argument to be omitted (needs new machinery possibly to allow optional. And then maybe allowmemory
to be provided as optional argument (falling back to env.memory) on recall and memorize using same machinery))The text was updated successfully, but these errors were encountered: