graphify
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Graphify β Claude Code uchun doimiy xotira π§
Claude Code'da har yangi chat ochganingizda loyihani noldan tushuntirasizmi? Graphify buni hal qiladi: barcha qarorlar, xulosalar va muhim ma'lumotlar kompyuteringizdagi lokal bazada saqlanadi, Claude har yangi chatda undan faqat keraklisini oladi. Natija: kam token sarfi, kam takrorlash, Claude eski qarorlaringizni "eslaydi".
Hammasi lokal ishlaydi β indekslash va qidirish 0 token, hech qanday API kalit yoki ro'yxatdan o'tish kerak emas. Ma'lumotlaringiz hech qayerga jo'natilmaydi.
Eslatma: bu loyiha GitHub'dagi mashhur
graphify(kod-bazani bilim grafiga aylantiruvchi CLI,Graphify-Labs/graphify) bilan hech qanday aloqasi yo'q β nom bir xil bo'lsa-da, vazifasi butunlay boshqa: bu β Claude Code uchun shaxsiy xotira MCP serveri.
Xususiyatlar
search_knowledge/save_memoryβ Claude o'zi vazifa boshida qidiradi, oxirida saqlaydi (qo'lda ham chaqirsa bo'ladi)Loyiha bo'yicha izolyatsiya β bir loyihaning xotiralari boshqasiga aralashmaydi (xohlasangiz hammasidan qidirsa bo'ladi)
Vizual dashboard (
/graphifyyokistart_dashboard.bat) β uchta ko'rinish:Ro'yxat β barcha xotiralarni ko'rish, qidirish, filtrlash, o'chirish
Graf β xotiralar orasidagi semantik o'xshashlik force-directed graf sifatida, loyiha bo'yicha rangli halqa + checkbox filter bilan
Modellar β embedding modelini tanlash (lokal/OpenAI/Ollama/maxsus API)
Token tejash statistikasi β
search_knowledgeqancha token tejaganining taxminiy hisobi (dashboard'da va/graphify tokenbilan)
Related MCP server: Cortex
Talablar
Windows 10/11
Python 3.10+ β o'rnatishda "Add python.exe to PATH" katagini albatta belgilang
Claude Code (desktop yoki CLI)
O'rnatish (2 daqiqa)
Bu repo'ni klonlang yoki ZIP qilib yuklab oling, doimiy joyga qo'ying (masalan
C:\graphifyβ keyin joyini o'zgartirmang!)install.batustiga ikki marta bosingTugagach Claude Code'ni yopib, qayta oching
Tamom. Tekshirish uchun chatda /mcp yozing β ro'yxatda graphify ko'rinishi kerak.
Qanday ishlaydi
Hech narsa qilishingiz shart emas β Claude o'zi:
vazifa boshida bazadan eski qarorlarni qidiradi (
search_knowledge)sessiya oxirida xulosani saqlaydi (
save_memory)
Qo'lda ham buyura olasiz:
"graphify'dan qidir: to'lov moduli qanday qilingan edi"
"bu qarorni graphify'ga saqla"
"graphify'dagi eslatmalarni ko'rsat"
/graphify slash-komandasi ham o'rnatiladi:
Komanda | Natija |
| Vizual dashboardni ochadi (http://localhost:5000) |
| Saqlangan xotiralar ro'yxati |
| Matnni xotiraga saqlaydi |
| Token tejash statistikasi (taxminiy) |
| Semantik qidiruv |
Xotiralar loyiha bo'yicha avtomatik ajratiladi β bir loyihaning qarorlari boshqasini chalg'itmaydi, lekin kerak bo'lsa hammasidan qidirsa bo'ladi.
Hujjatlarni qo'shish (ixtiyoriy)
.md/.txt hujjatlaringizni bazaga indekslash:
.venv\Scripts\python.exe indexer.py "C:\yo'l\hujjatlar-papkasi" --project LoyihaNomiEski loyihani tanishtirish
O'sha loyihada chat ochib, bir marta ayting:
"Loyihani o'rganib chiqib, asosiy arxitektura, texnologiyalar va muhim qarorlarni graphify'ga saqla"
Keyingi barcha chatlar shu tayyor xotiradan boshlanadi.
Embedding modelini almashtirish
Standart v1 (jonli) holatda β all-MiniLM-L6-v2 (384-d, lokal, bepul, internet/kalit shart emas). v2 branchida esa Ollama bge-m3 (1024-d) modeliga o'tish tayyorlangan. Dashboard'ning "Modellar" bo'limida OpenAI/Ollama/maxsus API'ga o'tsa bo'ladi β lekin bazada allaqachon ma'lumot bo'lsa, bu buzuvchi o'zgarish: eski vektorlar yangi model bilan mos kelmaydi. Amalda almashtirish uchun kg_db/ ni bo'shatib, indexer.py bilan qayta indekslash kerak. Batafsil: CUSTOM_MODELS.md va migrate_v2/README.md.
Tizim arxitekturasi: v1 va v2
Hozirgi jonli holat: v1 (jonli) β
all-MiniLM-L6-v2(384-d, gate 0.8) modelida barqaror ishlab turibdi β v2 branchda, switch kutadi.v2 yangilanishi (
feat/graphify-v2branchda):Embedding modeli: Ollama
bge-m3(1024-d, ko'p tilli β o'zbek va ingliz tillarida yuqori sifat).Model identity tekshiruvi: Baza metama'lumotlarida model nomi va o'lchami (
bge-m3, 1024-d) qat'iy tekshiriladi; agar Ollama yoki model mavjud bo'lmasa, qidiruv to'xtaydi (fail-closed, hech qachon noto'g'ri modelga o'tib ketmaydi).Leksik gate (0.5): So'rov so'zlarining kamida 50% qamrovi (
MIN_COVERAGE = 0.5) talab qilinadi (eski 0.8 dan tushirilgan, bu semantik moslikni to'sib qo'ymaydi).Cross-encoder reranker: Ixtiyoriy ONNX reranker qo'llab-quvvatlanadi, biroq resurs tejash va past kechikish uchun sukutda o'chiq (
GRAPHIFY_RERANK=0).To'liq o'tish va sinxronizatsiya yo'riqnomasi:
migrate_v2/README.md.
Maslahatlar
Sifat > miqdor: hamma narsani emas, qaror va xulosalarni saqlang
Eskirgan xotiralarni o'chirtirib boring: "mem_xxx ni o'chir"
Butun xotira
kg_db/papkasida β backup qilsangiz, hech narsa yo'qolmaydi
O'chirish
uninstall.bat ni ishga tushiring β Claude Code'dan uziladi. Keyin papkani qo'lda o'chirsangiz bo'ladi.
Muammolar
Muammo | Yechim |
"Python topilmadi" | Python'ni PATH bilan qayta o'rnating |
| Claude Code'ni to'liq yopib qayta oching |
Birinchi qidiruv sekin | Normal β model birinchi marta yuklanadi |
Papkani ko'chirdim, ishlamayapti |
|
Litsenziya
MIT β LICENSE fayliga qarang.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogβ¦
This server cannot be deployed
Maintenance
Related MCP Connectors
Persistent memory for AI assistants. Save once; recall from Claude, ChatGPT, or any MCP client.
Shared memory MCP server for Claude Code, Codex, Gemini CLI and Cursor. Cited, editable.
Persistent AI memory shared across Claude, ChatGPT, coding agents, and compatible MCP clients.
Persistent memory layer for AI tools. Save and recall notes across Claude and other MCP clients.
Related MCP Servers
- AlicenseAqualityDmaintenancePersistent memory MCP server that allows Claude to store, organize, and retrieve knowledge across sessions without consuming context window tokens.2482 PyPI18MIT
- AlicenseNot gradedqualityDmaintenancePersistent memory MCP server for Claude Code that captures and recalls project context across sessions, eliminating the need to re-explain architecture and decisions daily.1MIT
- AlicenseBqualityCmaintenanceA persistent memory MCP server for Claude Code that automatically saves conversations and retrieves relevant history across sessions to provide context.1729 PyPI10MIT
- AlicenseNot gradedqualityCmaintenanceA persistent memory MCP server for Claude Code that enables long-term recall across sessions via hybrid search, code intelligence, and tools for reading/writing memory.5MIT