issues Search Results · language:Dune language:Python language:JavaScript language:JavaScript language:Python language:PHP
Filter by
46.1M results
store.py s header says All SQL lives here. server.py contains zero SQL but get_token_stats, get_access_log,
get_prune_queue, resolve_prune_queue run raw SQL on _store._conn. Move those queries into MemoryStore ...
medium
Контекст
Категория «Активность» в GramGPT — действия, в которых аккаунт «реагирует» на контент (без публичных следов как у
реакций). Все 3 — read-equivalent для anti-spam Telegram, но watch_video и listen_voice ...
backend
enhancement
epic:advanced-warmup
p2
ready-for-agent
Counts come from grepping processed: N failed: M in stdout/stderr — fragile against any log format change. Have
watcher.py emit a final JSON summary line (or write a result file) and parse that. (PATH ...
medium
Compressed copies keep the original id with altered content — anything that round-trips content (re-add, export/import)
hashes differently and duplicates. Also the word-trim loop re-tokenizes the whole ...
medium
Each search result triggers boost_on_access with its own commit (5 commits/search). Batch into one executemany + single
commit. Design question: boosting on retrieval means frequently-searched-but-ignored ...
medium
Memories store %Y-%m-%d while access_log stores full ISO timestamps. find_stale_project_status compares created_at ?
lexically, so two entries on the same day never supersede. Store full ISO timestamps ...
medium
bandit is also run as a standalone in tox.ini.
flake8 --select S aiosmtpd
aiosmtpd/__init__.py:20:1: S101 Use of assert detected. The enclosed code will be removed when compiling to optimised byte code. ...
apply_decay (server.py), MemoryStore.auto_prune, and the prune dry-run each reimplement effective-score. They already
disagree: apply_decay clamps to min_score before importance/access boosts; auto_prune ...
medium
update_memory(facts=[...]) inserts new rows; it cannot modify existing ones. Clients calling it expecting mutation
create near-duplicates, which then feed the subset-pruner. Real gap: no MCP tool edits ...
medium
Problem
analytics.json (~100KB, growing) is fully reparsed and rewritten with indent=2 every 10th call, while everything else
moved to SQLite. _flush_analytics has except Exception: pass — failures are ...
high

Learn how you can use GitHub Issues to plan and track your work.
Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub IssuesProTip! Restrict your search to the title by using the in:title qualifier.