užklausiau ir apie tai, kaip protingiau reiktų migruot :)
Žinutės
Šis puslapis nėra tiesioginis — žinutės atnaujinamos pagal tvarkaraštį. Paskutinį kartą atnaujinta . Kito atnaujinimo laukiama apie . Laikas rodomas Vilniaus laiku (EET).
nu video gal nelabai suspaus jau 🙂
kurie ir taip spausti 😄
bet visokie kitokie dalykai tai buna ir 1.7x compression ratio pasiekia 😉
failu sistemos lygyje
tai tu tiesiog krauni ir matai kad kas su windows uzeme 1tb dabar uzima pvz 700tgb 😄
ir dar gali dakraut 🙂
spool/mempalacevec compressratio 1.80x -
spool/psql compressratio 1.48x -
rpool/webbackups compressratio 2.21x -
vo, tai jeigu tokie kaip webbackups, tia i 1tb nvme sukistu 2.2tb 😄
Done. Two surgical edits to M.md:
1. New standing section "Memory-first tool use" (after Memory-first exploration, M.md:149):
- mempalace_search reframed as mid-session reflex, not startup-only.
- Trigger = moment of doubt on a tool (MCP/skill/CLI/flag): recall params + calling pattern from palace before the
call.
- Calls out deferred-tool pattern explicitly — recall the ToolSearch: "select:<tool_name>" activation instead of
failing a direct call.
- Grep/schema-read only the gap palace didn't answer.
- Write back learned usage (links to Step 4).
2. Step 4 write-back bullet — new "Tool-usage notes" item: journal correct params/sequence/gotcha tagged by tool name
(TOOL:create_tunnel|...) so next session's recall query lands.
Loop closed: recall tool usage → fumble only the gap → write back how-to. Mirrors existing code-recall pattern, same
audit-gate philosophy. Grounded in palace hits (deferred-load idiom, mempalace usage notes).
Want me to also file this M.md change to the palace, or leave it?
1. New standing section "Memory-first tool use" (after Memory-first exploration, M.md:149):
- mempalace_search reframed as mid-session reflex, not startup-only.
- Trigger = moment of doubt on a tool (MCP/skill/CLI/flag): recall params + calling pattern from palace before the
call.
- Calls out deferred-tool pattern explicitly — recall the ToolSearch: "select:<tool_name>" activation instead of
failing a direct call.
- Grep/schema-read only the gap palace didn't answer.
- Write back learned usage (links to Step 4).
2. Step 4 write-back bullet — new "Tool-usage notes" item: journal correct params/sequence/gotcha tagged by tool name
(TOOL:create_tunnel|...) so next session's recall query lands.
Loop closed: recall tool usage → fumble only the gap → write back how-to. Mirrors existing code-recall pattern, same
audit-gate philosophy. Grounded in palace hits (deferred-load idiom, mempalace usage notes).
Want me to also file this M.md change to the palace, or leave it?
- ❤2
suku galvą kaip pasiruošt migracijai, kas svarbu, kas ne ir ką reik kardinaliai persidaryt arba jau migravus persidaryt (vis girdžiu go lang.. gal net dar keletą savaičių sudeginti planus, tam pokyčiui, nes belenkaip pabodo nestabilumas sistemos; overkill ryškiai pas mane ir nebesusigaudo pats opusas ko pridaryta) :D
servo ar personal laptopo?
siandien jau kazkaip sugalvojo viduri darbo playwright is mempalace atsimint 😄
every day something new!
kaip naudot atsimine* 🙂
- 😁1
as su 250gb vargstu in total 😄
uztai visus memory ir ka galiu - laikau remote serve 😉
- ❤1
jis paieskojo mempalace
"overkill " siandien
bet po pusmecio visu projektu atmintis
nauja startuoji jis pasiima know how is kito wing
nebereik isradinet dviracio
pradejau uzvakar mini "api" kitam projektui parasiau prompt
sure, why not 😄
- 🔥1
bet galvoju i golang perrasyt
pain tas python, su golang butu 0 dependencies
1 binary failas -mcp serveris gal net remote
SaaS memory 😄
suvedi mcp savo user/pass ir ... veikia tiesiog atmintis uz 5$ per men :))))
😄
https://github.com/MemPalace/mempalace/issues paskutiniai 5 issue mano
merge wings dasidejau , recompute tunnels and dar kazka ten po merge
ollama pasidariau i rnaudoju bge-m3 modeli embedingui ne ta kuris buvo default
sumamrizink ka perdarei?
Rodoma paskutinė surinkta leidžiama būsena; pašalinimų suderinimas nebaigtas. Visas pokalbis — Telegram grupėje.