Yakha futhi Wethule Iqoqo Lezindawo ze-AI Eligxile Emagameni Angukhiye nge-shipwords

Okwenziwa yi-shipwords

shipwords iwuhlaka oluvulekile olusebenzisa ilayisensi ye-MIT lokusebenzisa iqoqo lamasayithi ezithombe namavidiyo e-AI ku-Cloudflare, isayithi ngalinye linikezelwe egameni elingukhiye elilodwa. Amakhono abiwe afana nama-akhawunti, izinkokhelo, amakhredithi, nokukhiqizwa kwe-AI asetshenziswa kanye kuphela kuhlaka, kuyilapho isayithi ngalinye lingaba namakhasi alo, ubuwena obubonakalayo, isizinda, isizindalwazi, isitoreji, ne-akhawunti yokukhokha.

Isayithi ngalinye liwuhlelo lwe-TanStack Start ne-React 19 olusebenza ku-Cloudflare Worker yalo, lusebenzisa i-D1, i-R2, nama-Workflows. Lokhu kwehlukaniswa kukuvumela ukuthi uqalise, ubuyekeze, noma uyeke amasayithi ngokuzimela, kuyilapho usathola ukuthuthukiswa komgogodla owabiwe ekukhishweni okuzayo.

Izici ezibalulekile

• Ukungena ngemvume nge-Google nange-imeyili, kanye namakhasi e-akhawunti.
• Ukusekelwa kwe-Stripe ne-PayPal, okuhlanganisa amaphakheji amakhredithi, okubhaliselwe, ama-oda, ukubuyiselwa kwemali, nezingxabano.
• Izaziso ze-imeyili kanye nokuzama kwamahhala kwezivakashi, okunemikhawulo yokusetshenziswa kabi ye-reCAPTCHA v3.
• Amakhasi obumfihlo, emigomo, okubuyiselwa kwemali, ne-DMCA.
• Amathuluzi e-SEO ne-GEO, okuhlanganisa sitemap.xml, robots.txt, kanye llms.txt.
• Isikhathi sokusebenza sokukhiqiza esisekelwe kuma-workflow, esinezimo zemisebenzi eziya ohlangothini olulodwa kanye nokubuyiselwa kwemali okuzenzakalelayo kwemisebenzi ehlulekile.
• Ukuhlanganiswa kwe-Yir API kwamamodeli afana ne-GPT Image, Nano Banana, Seedream, FLUX, Seedance, Veo, Kling, ne-Wan.
• I-mock upstream yokuthuthukisa kwasendaweni ngaphandle kokhiye we-model API.
• Ukuhlanganiswa kwamavidiyo ngokuzithandela ngesiqukathi se-ffmpeg.

Okudingekayo ngaphambi kokuqala

Ngaphambi kokuqala, faka i-Node 24, njengoba kuchazwe yi-.nvmrc yenqolobane, bese unika amandla i-Corepack ukuze inguqulo elindelekile ye-pnpm itholakale:

corepack enable

Ku-Windows, sebenzisa i-Git Bash noma i-WSL. I-akhawunti ye-Cloudflare kanye ne-akhawunti ye-Stripe noma ye-PayPal kuyadingeka uma usulungele ukukhipha isayithi futhi wamukele izinkokhelo. Akudingeki ekusethweni okuyisisekelo kwasendaweni.

Faka inqolobane

Yenza i-fork noma u-clone wenqolobane, bese ufaka okuncikile kuyo:

pnpm install

Yakha isayithi elisha

Sebenzisa umakhi wamasayithi ukuze wakhe isayithi usebenzisa ithempulethi ephelele yesayithi le-AI. Impikiswano yokuqala iba yinkomba engaphansi kwe-sites/, kanti isizinda sigcinwa ekucushweni kwesayithi:

pnpm new-site my-tool --domain my-tool.com
pnpm install

Isayithi elakhiwe lihlanganisa ikhasi lasekhaya, amanani, ukuhamba kwe-akhawunti nokungena ngemvume, amakhasi ezinqubomgomo, kanye namafayela ajwayelekile e-sitemap.xml, robots.txt, ne-llms.txt. Phinda usebenzise pnpm install ukuze kufakwe okuncikile kwesayithi elisha.

Lungisa izimfihlo zendawo

Gcina izimfihlo zendawo ngaphandle kwenqolobane. Dala inkomba yokucushwa, kopisha ifayela lezinguquko eziyisibonelo lesayithi elakhiwe, bese uxhumanisa ukucushwa kwe-worktree:

mkdir -p ~/.config/shipwords
cp sites/my-tool/.dev.vars.example ~/.config/shipwords/my-tool.dev.vars
pnpm setup:worktree

Ukucushwa okuyisibonelo kuklanyelwe ukusebenza ngokushesha ekuthuthukiseni kwasendaweni. Amakhodi okungena ngemvume nge-imeyili aphrintwa kutheminali yeseva yokuthuthukisa, ama-akhawunti amasha athola amakhredithi okubhalisa, futhi ukukhiqiza kusebenzisa i-mock upstream uma UPSTREAM_MOCK=1. Lokhu kusho ukuthi ungahlola uhlelo ngaphandle kokhiye we-model API.

Engeza ukukhiqizwa kwangempela

Ngezicelo zamamodeli angempela, engeza ukhiye we-Yir API bese ucisha i-mock upstream ngokushintsha UPSTREAM_MOCK.

Hlola izinkokhelo

Ukuhlola ukuthengwa kwamakhredithi, engeza ukhiye wokuhlola we-Stripe oqala ngo-sk_test_. Kusuka enkombeni yesayithi, sebenzisa ithuluzi lenqolobane lokudlulisela ama-webhook ukuze ukuthuthukisa kwasendaweni kwamukele imicimbi ye-Stripe:

node ../../tools/stripe-listen.mjs

Ikhodi yokukhokha eyabiwe yakhelwe ukuphatha imicimbi yokukhokha ephindaphindiwe noma engahlelekile, ukukhokhiswa kabili, kanye nokubuyiselwa kwemali kwemisebenzi ehlulekile.

Lungisa ukungena ngemvume kokuzithandela nokufinyelela kwezivakashi

Ukuzama kwamahhala kwezivakashi kudinga okhiye be-reCAPTCHA. Ngaphandle kwabo, izivakashi kufanele zingene ngemvume ngaphambi kokukhiqiza kwazo kokuqala. Ukungena ngemvume nge-Google kuyazikhethela; uma kunikwe amandla, dala iklayenti le-Google OAuth bese usebenzisa http://localhost:<port> njengomsuka ovunyelwe. Ukungena ngemvume nge-imeyili kusebenza ngaphandle kwe-Google OAuth.

Qalisa isayithi endaweni

Yiya esayithini elakhiwe, dala izinto zesizindalwazi zasendaweni, sebenzisa ukufuduka kwasendaweni, bese uqala iseva yokuthuthukisa:

cd sites/my-tool
pnpm db:generate && pnpm db:migrate:local
pnpm dev

Vula i-URL yendawo ephrintwe yiseva yokuthuthukisa. Ungahlola ukuqinisekiswa nge-imeyili, ukwakhiwa kwe-akhawunti, amakhredithi, nokukhiqiza kwe-mock usebenzisa ukucushwa okuyisibonelo.

Sebenzisa imiyalo yenqolobane

Imiyalo eminingi ingagxiliswa kusayithi elilodwa noma isetshenziselwe ukuhlola yonke inqolobane:

pnpm new-site <dir> --domain <domain>
pnpm list-sites
pnpm setup:worktree
pnpm --filter "./sites/<site>" dev
pnpm --filter "./sites/<site>" check
pnpm check:packages
pnpm check:template
pnpm lint && pnpm test:tools

Umyalo we-list-sites wakha uhlu lwamasayithi usebenzisa ikhodi ne-README yesayithi ngalinye. Umyalo we-check wesayithi wenza ukuhlola izinhlobo, ukwakha, nokuhlola ukuqalisa. Ukuhlolwa kwamaphakheji kuhlanganisa umgogodla namaphakheji emisebenzi okwabelwana ngawo, kuyilapho ukuhlolwa kwethempulethi namathuluzi kusiza ukubamba izinkinga ngaphambi kokwethulwa.

Yakha isayithi nge-ejenti yokubhala ikhodi ye-AI

Uma usulazi igama elingukhiye ofuna ukulikhomba, inqolobane isekela ukuhamba komsebenzi okuqhutshwa yi-ejenti. Ku-Claude Code, Codex, noma Cursor, nikeza imiyalo ye-ejenti yenqolobane bese uyicela ukuthi yakhe isayithi:

Read shipwords.dev/llms.txt and build a site for "<your keyword>"

I-agent ingakwazi uku-clone-a i-repository, yakhe isayithi, futhi iyiqhube endaweni yasendaweni. Imithetho yama-agent okubhala ikhodi nge-AI ibhalwe ku-AGENTS.md. Uma udinga imibono yamagama angukhiye asafufusa, iphrojekthi ehlukile ye-new-term-radar iqapha amagama amasha e-AI ku-X, Hugging Face, GitHub, nakuma-sitemap.

Lungiselela ukuziphatha kwesayithi okuthuthukile

Sebenzisa i-runtime yokukhiqiza eyabiwe

Iphakheji ye-packages/jobs inikeza i-runtime esekelwe ku-workflow, i-interface ye-upstream adapter, i-Yir adapter, i-local mock upstream, kanye ne-ffmpeg composition ongayikhetha. Gcina ukuziphatha kokukhiqiza okuqondene nesayithi ku-configuration yesayithi, futhi usebenzise i-runtime eyabiwe ekucubunguleni imisebenzi nasekuphatheni ukubuyiselwa kwemali.

Setha imikhawulo yezindleko zokukhiqiza

Izingcingo zamamodeli nge-Yir zikhokhiswa ngesicelo ngasinye. Izindleko eziphezulu zomsebenzi ngamunye zisethwa ku-site.models.ts. Buyekeza le mingcele ngaphambi kokuvumela ukukhiqiza kwangempela ukuze amanani esayithi nokuziphatha kwama-credit kuhambisane nezindleko eziphezulu ze-upstream.

Funda isibonelo sevidiyo

Esayithini egxile kuvidiyo, hlola i-sites/tinydeskai-example. Lona ngumthombo ophelele we-Tiny Desk AI futhi ubonisa izikhala zezithombe, ukuhlela ividiyo ngohambo olulodwa, ukuhlola okukhiphayo, ukuhlanganisa nge-ffmpeg, namakhasi okuqukethwe. I-workflow yayo iguqula isiqeshana esisabalele sibe yithempulethi izivakashi ezingayisebenzisa ngezithombe ezilayishiwe.

Gcina amasayithi ezimele

Isayithi ngalinye lithunyelwa ngokuzimela, line-database, isitoreji, kanye ne-akhawunti yokukhokha yalo. Le ndlela yokwakha ikuvumela ukuthi uthuthukise i-core eyabiwe kanye, bese uthumela isibuyekezo kumasayithi akhethiwe ngaphandle kokuxhumanisa ukwethulwa noma ukuhoxiswa kwawo.

Lungiselela ukukhiqizwa

Uma isayithi lasendaweni selisebenza, landela incwadi yokuqalisa isayithi ngenqubo yokukhiqiza. Hlela izindleko zesizinda, ukusetshenziswa kwe-Cloudflare, izimali zokucubungula izinkokhelo, nezicelo zamamodeli e-Yir. Amasayithi evidiyo ahlanganisa iziqeshana asebenzisa i-Cloudflare Containers futhi adinga uhlelo lwe-Workers Paid.

Ngaphambi kokuqalisa, qinisekisa ukuqinisekiswa kwesayithi, ama-webhook okukhokha, ukuziphatha kwama-credit, ukwehluleka kokukhiqiza, ukubuyiselwa kwemali, amakhasi ezinqubomgomo, amafayela e-SEO, kanye nokucushwa kwe-analytics. I-repository iphinde inikeze imibhalo ye-architecture kanye nokuthuthukiswa kwasendaweni ukuze uthole imininingwane ejulile yokucushwa.

Isiphetho

i-shipwords inikeza isisekelo esisebenzayo sokwethula amasayithi amaningi e-AI agxile kumagama angukhiye ngaphandle kokwakha kabusha ukuqinisekiswa, ukukhokhisa, ama-credit, ama-workflow okukhiqiza, nokusekelwa kwe-SEO kuyo yonke iphrojekthi. Qala ngethempulethi, sebenzisa i-mock upstream ekuhloleni kwasendaweni, engeza abahlinzeki bangempela uma usulungile, bese uthumela isayithi ngalinye eliqediwe ngokuzimela ku-Cloudflare.