Add house-style memory agent (SuperDocs use-case) - #55
Open
suhitk1 wants to merge 1 commit into
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add house-style memory agent (SuperDocs use-case)
Built by Suhit Kumar for the SuperDocs Round 2 candidate task.
An agent that edits documents through SuperDocs and, with consent, keeps a durable per-customer memory of corrections received — so it needs fewer corrections on the same class of document over repeated sessions.
The deliverable is a measured curve, not just the agent: correction count per document across 20 documents, memory off vs. memory on, with a live noise-floor control (memory-off run twice). Result: mean 17.7 (off) / 17.0 (off repeat) / 7.95 (on) — an effect of 9.4, roughly 2.5× the noise floor. Also includes a cross-customer leak test (two opposed fictional client style guides, zero cross-client marker leakage net of baseline) and a wipe test showing real reversion after clear_cross_session_memory.
Full methodology, raw run data, and an honest limitations section (including where rigor was deliberately reduced under time constraints) are in the README.