Nangabe inkampani yakho ine-app ye-AI, i-chatbot nobe lithulusi langekhatsi lelixhumana nemamodeli lanjenge-GPT, Claude nobe Gemini, buta licembu lakho letebuchwepheshe umbuto munye namuhla: ingabe kukhona iseva, i-laptop nobe i-pipeline yekwakha leyafaka i-LiteLLM inguculo 1.107.0 mhla ti-21 Imphala 2026? Lenguculo yashicilelwa ngabahlaseli futsi yahlala ku-PyPI cishe emahora lamabili, emkhatsini wa-14:39 na-16:23 ngesikhatsi saseRiyadh, yadawunilowudwa cishe tikhatsi letingu-49,000. Yeba tikhiye te-cloud, tikhiye tebahlinzeki be-AI, emaphasiwedi ema-database netikhiye te-SSH. Nangabe imphendvulo itsi yebo, kuyisusa akwanele: phindze wakhe umshini kusuka ku-image lehlantekile futsi ushintje sonkhe sikhiye lebesifinyeleleka kuwo. Inguculo lephephile ngu-1.106.2.

Kwenteke ini ngco?

I-LiteLLM ngumtapo wetinhlelo ne-gateway lokuvulekile umtfombo levumela app yinye kutsi ikhulume nemamodeli e-AI lamanyenti ngekusebentisa sixhumi sinye, kungako tinkampani letinyenti tiyibeka enhliziyweni yemikhicito yato ye-AI. Ngekusho kwe-Chainguard, yadawunilowudwa tikhatsi letingetulu kwa-100 million ku-PyPI enyangeni lendlulile, futsi emaphakheji langetulu kwa-2,000 ancike kuyo, ngako emacembu lamanyenti ayifaka angati.

Ngekusho kwembiko losemtsetfweni wesehlakalo we-LiteLLM, bahlaseli batfola ema-token ekungena e-GitHub labekaphume kuma-log ephrojekthi. Mhla ti-21 Imphala nga-10:08 UTC basebentisa imvume yemtfutfukisi lonemalungelo lakhetsekile kutsi beca kuvikelwa kwe-branch futsi bashintja indlela yekukhipha kuze yakhe kusuka endzaweni lenelibito lelifanako, «Berri-AI/litellm» esikhundleni se-«BerriAI/litellm». Nga-11:39 UTC emafayela lamabi afika ku-PyPI, futsi nga-11:46 UTC kwaphuma ema-image e-Docker langcolile nema-tag la-v1.107.0, latest, main-latest na-main-stable. I-PyPI yasusa lenguculo ngaphambi kwa-13:23 UTC, futsi ema-tag e-Docker ahlantwa nga-14:20 UTC, lokusho nga-17:20 eRiyadh.

Lokhu kuhlasela kwesibili kulephrojekthi lenye nyakenye. I-Chainguard iphawula kutsi ngeNdlovana 2026 licembu lasebentisa i-token ye-PyPI lebelwebiwe kushicilela tinguculo 1.82.7 na-1.82.8.

Yeba ini lenguculo lembi?

Lokuyingoti kutsi ayilindzi kutsi app yakho isebentise i-LiteLLM. Lenguculo yengeta emafayela ekucala e-Python lasebenta njalo nangabe i-Python icala kulowo mshini, ngisho nakumyalo longahlobani nayo. Ngekusho kwekuhlatiya kwe-Wiz, lelisela labutsa loku:

  • Tikhiye te-cloud te-AWS, Google Cloud ne-Azure kusuka kumafayela ekuhlela, kuma-variable endzawo nakumininingwane yeseva.
  • Tikhiye te-API te-OpenAI, Anthropic, Cohere nalabanye bahlinzeki be-AI, lokusho sikweleti sakho nemininingwane yakho.
  • Tikhiye te-SSH tangasese, umlandvo we-shell, kanye nemaphasiwedi nema-cookie lagcinwe kuma-browser.
  • Kufika kuma-500 emvume yebahlinzeki legcinwe ku-database ye-LiteLLM Proxy.

Bahlaseli abazange bahlale nalokwebiwe. I-Wiz yabona tikhiye te-AWS letebiwe tilingwa cishe emahora langu-4.5 ngemuva kwekwebiwa, kusuka kumaseva ekwakha nakuma-node e-Kubernetes. Lohlelo lolubi lwaphindze lwafaka sigaba sesibili lesitsatsa ikhodi lekhashane njalo ngelihora, ngako umshini uhlala uvulekile ngisho nangemuva kwekususwa kweliphakheji.

Ingabe inkampani yami itsintsekile?

Mhlawumbe cha, nangabe licembu lakho libeka tinguculo letinembile futsi alizange libuyekete lutfo emkhatsini wa-14:39 na-17:20 ngesikhatsi saseRiyadh mhla ti-21 Imphala. Ungatsintseka nangabe kwenteke loku lokulandzelako:

  • Iseva nobe i-container isebentisa «pip install litellm» ngaphandle kwenguculo lebekiwe, nobe idonsa i-tag ye-Docker «latest».
  • I-pipeline ye-CI/CD yaphindze yakha ngalentsambama.
  • Umtfutfukisi walinga i-LiteLLM ku-laptop yakhe ngalelo langa.
  • Luhlelo lwakho lusebentisa leliphakheji lelinye lelincike ku-LiteLLM.

I-LiteLLM itsi emakhasimende enguculo yayo yangasese ye-Enterprise awazange atsintseke.

Tinyatselo letinhlanu tanamuhla

  • Funa: inguculo 1.107.0 kumafayela e-lock, kuma-log ekwakha nakuma-image e-container, kanye nemafayela i-cook.pth ne-sitecustomize.py kufolda yemaphakheji e-Python.
  • Phindze wakhe: nangabe uyitfola, unganeliswa kuyisusa kuphela. Umbiko losemtsetfweni uyecwayisa kutsi kusebentisa i-pip endzaweni lenegciwane kungaphindze kucalise lohlelo lolubi. Phindze wakhe umshini nobe i-container kusuka ku-image lehlantekile.
  • Shintja tonkhe tikhiye: tikhiye nemisebenti ye-cloud, tikhiye tebahlinzeki be-AI, emaphasiwedi ema-database netikhiye te-SSH lebetifinyeleleka kulomshini. Cabanga kutsi sesisetandleni tebahlaseli.
  • Buketa ema-log: hlola umsebenti we-cloud kusukela mhla ti-21 Imphala ufuna tincingo letingakavami letinjenge-GetCallerIdentity, futsi ubuke sikweleti somhlinzeki wakho we-AI ufuna kusetjentiswa kwemali longakwati.
  • Beka tinguculo: faka i-litellm==1.106.2 nobe kulungiswa lokusemtsetfweni lokulandzelako, futsi ubeke onkhe umtapo wetinhlelo ngefayela le-lock esikhundleni sekudonsa «latest» ngalokutentakalelako.

Nangabe usifundzile sihloko setfu sangaphambilini mayelana netikhiye letikhohlakele ekhodini yema-agent e-AI, sifundvo siyafana ngelinye luhlangotsi: tikhiye te-AI manje setiphakathi kwetimfihlo letiligugu kakhulu tenkampani yakho, futsi indzawo letiphatsiswe kuyo iba yinhloso.

Umbono we-Origami

Kute lobegebenge ikhodi ye-LiteLLM ngokwayo lapha. Bahlaseli bashintja indlela leyakhiwa futsi ishicilelwe ngayo, ngako kubuketa ikhodi yemtfombo kuphela bekungeke kukubambe loku. Kungako kuvikela imikhicito ye-AI akuncikanga kakhulu ekukhetseni umtapo wetinhlelo, kodvwa endleleni lowusebentisa ngayo: tinguculo letibekiwe, sikhiye lesehlukile endzaweni ngayinye lesinemvume ledzingekako kuphela, imikhawulo yekusebentisa imali kumhlinzeki ngamunye we-AI, kanye nerekhodi lelicacile lekutsi ngubani lofinyelela kuphi. Nasakha tinhlelo te-AI tetinkampani e-Origami, siphatsa i-gateway yemamodeli njengencenye lebucayi kakhulu yeluhlelo, ngobe iphetse tikhiye tako konkhe. Nangabe ufuna kutsi i-app yakho ye-AI ihlolwe mayelana nalesehlakalo, nobe kuhlelwa kabusha kwetikhiye nemvume yayo, sitsintse.

Imitfombo