Faka futhi Usebenzise i-Open WebUI ye-AI Ebanjwe Wena

Yini i-Open WebUI?

I-Open WebUI iyinkundla ye-AI enwebekayo, ozibambele yona, eyakhelwe ukusebenza ngokuphelele ngaphandle kwe-inthanethi uma ilungiselelwe ngezinsiza zasendaweni. Ihlinzeka ngesixhumi esibonakalayo esisekelwe esipheqululini samamodeli e-Ollama asendaweni, ama-API ahambisana ne-OpenAI, kanye nenjini eyakhelwe ngaphakathi yokwenza i-retrieval-augmented generation, evame ukubizwa nge-RAG.

Ungayisebenzisa njengendawo yakho yomuntu siqu yokusebenza nge-AI noma uyifakele iqembu. Ngaphezu kwezingxoxo, le nkundla ihlanganisa amanothi, iziteshi ezabiwe, inkumbulo eqhubekayo, amakhalenda, ama-automation ahleliwe, ukuhlanganiswa kokukhiqizwa kwezithombe, izingcingo zezwi nevidiyo, izibalo, ukuhlolwa kwamamodeli, kanye nokulawulwa kokufinyelela okunemininingwane.

Izici Eziyinhloko

• Ukuxhumeka kwamamodeli okuguquguqukayo: Sebenzisa amamodeli e-Ollama asendaweni kanye nezinsizakalo ezihambisana ne-OpenAI ezifana ne-LMStudio, GroqCloud, Mistral, OpenRouter, ne-vLLM.
• I-RAG yasendaweni: Engeza amadokhumenti ezingxoxweni noma thola amafayela emtatsheni wolwazi usebenzisa umyalo we-#. Le nkundla isekela ukusesha okuhlanganisiwe, ukuhlela kabusha imiphumela, izinjini eziningi zokukhipha okuqukethwe, kanye nemininingwane yolwazi eyisishiyagalolunye esebenzisa ama-vector.
• Amamodeli nama-agent: Hlanganisa imodeli eyisisekelo nemiyalelo yangokwezifiso, ulwazi, amathuluzi, okuguquguqukayo okuguqukayo, kanye nemithetho yokufinyelela yomsebenzisi noma yeqembu.
• Ukweluleka ukusebenza: Engeza ama-Filters, ama-Actions, ama-Pipes, ama-Tools, nama-Skills, noma xhuma amaseva amathuluzi e-MCP, MCPO, ne-OpenAPI.
• Ukusebenzisana: Sebenzisa amanothi, iziteshi zesikhathi sangempela, imicu yezingxoxo, ukusabela, okuphiniwe, amakhalenda abiwe, kanye nokuhlelwa okusizwa yi-AI.
• Izindlela eziningi zokuxhumana: Lungiselela ukuguqulwa kwenkulumo ibe umbhalo, umbhalo ube inkulumo, izingcingo zezwi nevidiyo, kanye nezinjini zokukhiqiza noma zokuhlela izithombe.
• Ukuphatha: Lawula izindima, amaqembu, izimvume, ukuqinisekiswa, izibalo zokusetshenziswa, ukuhlolwa kwamamodeli, isitoreji, kanye nokuqapha ukusebenza.
• Ukuguquguquka kokufakwa: Faka usebenzisa i-pip, uv, Docker, Docker Compose, Kubernetes, Kustomize, noma Helm.
• Isixhumi esibonakalayo esizivumelanayo: Sebenzisa i-Open WebUI kukhompyutha noma kumakhalekhukhwini, noma uyifake njenge-Progressive Web App.

Khetha Indlela Yokufaka

Ukuze wenze ukuhlola okusheshayo kwasendaweni, sebenzisa i-pip noma i-Docker. I-pip isebenzisa i-Open WebUI ngokuqondile nge-Python, kanti i-Docker ihlinzeka ngokufakwa okuhlukanisiwe futhi yenza kube lula ukukhetha izithombe ezijwayelekile, ezinikwe amandla yi-CUDA, noma ezine-Ollama ngaphakathi.

Okubalulekile: Wonke umyalo we-Docker kufanele ufake i-open-webui:/app/backend/data. Le volumu ephikelelayo ivikela isizindalwazi kanye nenye idatha yohlelo lokusebenza lapho isitsha sishintshwa noma sisuswa.

Inketho 1: Faka nge-Python pip

I-Open WebUI idinga i-Python 3.11 ekufakweni nge-pip okuchazwe iphrojekthi. Qinisekisa ukuthi indawo yakho isebenzisa inguqulo elungile ye-Python ngaphambi kokuqhubeka.

1. Faka iphakheji:

   pip install open-webui

2. Qalisa iseva:

   open-webui serve

3. Vula http://localhost:8080 esipheqululini sakho.

Inketho 2: Sebenzisa i-Open WebUI nge-Docker

Xhuma ku-Ollama ekhompyutheni engumsingathi

Uma i-Ollama isivele isebenza kukhompyutha efanayo ne-Docker, qala isithombe esijwayelekile se-Open WebUI bese wengeza imephu yomsingathi ukuze isitsha sikwazi ukufinyelela kumsingathi:

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Ngemva kokuqala kwesitsha, vakashela http://localhost:3000.

Xhuma ku-Ollama kwenye iseva

Setha i-OLLAMA_BASE_URL ekhelini lesevisi ye-Ollama ekude:

docker run -d -p 3000:8080 -e OLLAMA_BASE_URL=https://example.com -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Faka i-https://example.com esikhundleni se-URL yangempela yeseva ye-Ollama.

Nika amandla ukwesekwa kwe-Nvidia GPU

I-Open WebUI ihlinzeka ngesithombe esimakwe nge-CUDA. Faka i-Nvidia CUDA container toolkit ku-Linux noma ku-WSL ngaphambi kokusisebenzisa:

docker run -d -p 3000:8080 --gpus all --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:cuda

Sebenzisa i-OpenAI API kuphela

Uma ungayidingi i-Ollama, dlulisela ukhiye wakho we-OpenAI API esithombeni esijwayelekile:

docker run -d -p 3000:8080 -e OPENAI_API_KEY=your_secret_key -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Faka ubufakazi obufanele esikhundleni se-your_secret_key. I-Open WebUI ingaxhuma nakwamanye ama-API ahambisana ne-OpenAI, ngokulungiselelwa okuqondene nomhlinzeki okuchazwe iphrojekthi.

Sebenzisa i-Open WebUI ne-Ollama Esitsheni Esisodwa

Isithombe se-:ollama sihlanganisa i-Open WebUI ne-Ollama. Sisebenzisa ivolumu eyodwa yamamodeli e-Ollama nenye yedatha yohlelo lokusebenza lwe-Open WebUI.

I-Ollama ehlanganisiwe ene-GPU

docker run -d -p 3000:8080 --gpus=all -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:ollama

I-Ollama ehlanganisiwe esebenzisa i-CPU kuphela

docker run -d -p 3000:8080 -v ollama:/root/.ollama -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:ollama

Vula http://localhost:3000 ngemva kokuqala kwesitsha.

Ukusetshenziswa Okuyisisekelo

1. Vula isixhumi esibonakalayo: Sebenzisa i-port 8080 ekusetheni kwe-pip noma i-port 3000 emiyalweni ejwayelekile ye-Docker engenhla.

2. Sebenzisa imodeli etholakalayo: Qala ingxoxo ngemodeli evezwa uxhumano lwakho lwe-Ollama noma oluhambisana ne-OpenAI olulungiselelwe.

3. Bhala imiyalelo enothile: Izimpendulo zisekela i-Markdown ne-LaTeX, okwenza isixhumi esibonakalayo sifanele izincazelo ezihlelekile, ikhodi, amathebula, kanye nezimpawu zezibalo.

4. Qhubeka kuwo wonke amadivayisi: Isixhumi esibonakalayo esizivumelanisayo sisebenza kumakhompyutha nakomakhalekhukhwini, kanti ukusekelwa kwe-Progressive Web App kunikeza ulwazi olufana nolohlelo lokusebenza.

Sebenzisa amadokhumenti nge-RAG

Faka amadokhumenti ngqo engxoxweni noma uwathole emtatsheni wakho usebenzisa umyalo othi #. I-Open WebUI ingakhipha futhi ikhombe okuqukethwe, ithole izindima ezifanele, bese izifaka kumongo wemodeli. Isitaki sayo se-RAG sisekela ukusesha okuhlanganisiwe kwe-BM25 namavekhtha, ukuhlelwa kabusha kwemiphumela, kanye nemodi yomongo ophelele.

Ekusethweni okukhulu, izinketho ezisekelwayo zemininingwane yamavekhtha zifaka i-ChromaDB, PGVector, Qdrant, Milvus, Elasticsearch, OpenSearch, Pinecone, S3Vector, ne-Oracle 23ai.

Faka iwebhusayithi engxoxweni

Sebenzisa umyalo othi # ulandelwe yi-URL ukuze udonsele okuqukethwe kwewebhusayithi engxoxweni. I-Open WebUI ingalungiselelwa futhi ukuze ivumele amamodeli ukuthi alande okuqukethwe kwewebhu lapho kudingeka.

Qhathanisa amamodeli amaningi

Izingxoxo zamamodeli amaningi zikuvumela ukuthi uxhumane namamodeli amaningi ngesikhathi esisodwa. Lokhu kuyasiza uma ufuna ukuqhathanisa ukucabanga, isitayela sokubhala, noma indlela abahlinzeki abasebenza ngayo. Abaphathi bangasebenzisa nenkundla eyakhelwe ngaphakathi, ukuhlolwa kwe-A/B, namabhodi abaphambili asekelwe ku-ELO ukuze bahlole amamodeli.

Amathiphu Okucushwa Okuthuthukile

Dala amamodeli nama-agent akhethekile

Hlanganisa imodeli eyisisekelo nemiyalelo yangokwezifiso, amathuluzi, nolwazi ukuze wakhe i-agent ekhethekile. Okuguquguqukayo okuguqukayo nezilawuli zokufinyelela zomsebenzisi ngamunye noma zeqembu ngalinye kwenza kube nokwenzeka ukulungisa i-agent ukuze ihambisane namaqembu ahlukene. Ama-preset omphakathi nawo angangeniswa ku-Open WebUI Community.

Nweba inkundla

I-Open WebUI isekela ama-Filters, ama-Actions, ama-Pipes, ama-Tools, nama-Skills. Izinsizakalo zangaphandle zingaxhunywa nge-MCP, MCPO, namaseva wamathuluzi e-OpenAPI. Lezi zindawo zokunweba zingasekela ukuhlanganiswa kwangokwezifiso, izinqubo zokugunyaza, imikhawulo yezinga, kanye nokuxhumana okwengeziwe kwedatha.

Hlela ukusatshalaliswa kwamaqembu namabhizinisi

Ezimweni zabasebenzisi abaningi, sebenzisa izindima, amaqembu, nezimvume ukulawula ukufinyelela. I-Open WebUI isekela ukuhlanganiswa kwe-LDAP ne-Active Directory, ukungena okukodwa okusekelwe kuma-header athembekile ne-OAuth, kanye nokuhlinzekwa kwe-SCIM 2.0 kubahlinzeki bobunikazi abafana ne-Okta, Azure AD, ne-Google Workspace.

Isitoreji singasebenzisa i-SQLite, ngokuzikhethela ngokubethela, noma i-PostgreSQL. Amafayela angagcinwa endaweni noma ku-S3, Google Cloud Storage, noma Azure Blob Storage. Amaseshini asekelwe ku-Redis nokusekelwa kwe-WebSocket kwenza kube nokwenzeka ukusatshalaliswa kwabasebenzi abaningi namanodi amaningi ngemuva kwama-load balancer, kanti i-OpenTelemetry inikeza ama-trace, amamethrikhi, namalogi.

Sebenza ungaxhunyiwe ngokuphelele

Lapho usebenzisa i-Open WebUI endaweni engaxhunyiwe, setha HF_HUB_OFFLINE ku-1 ukuze uvimbele imizamo yokulanda amamodeli ku-inthanethi:

export HF_HUB_OFFLINE=1

Ukusatshalaliswa okungaxhunyiwe kusadinga ukuthi wonke amamodeli adingekayo, okuncikile, nezinsizakalo zasendaweni kutholakale kuleyo ndawo.

Zama isithombe sokuthuthukisa ngokuqapha

Isithombe se-:dev siqukethe izinguquko ezingazinzile nezintsha kakhulu ezingase zibe namaphutha noma izici ezingakapheleli. Sisebenzise kuphela uma wamukela leyo ngozi:

docker run -d -p 3000:8080 -v open-webui:/app/backend/data --name open-webui --add-host=host.docker.internal:host-gateway --restart always ghcr.io/open-webui/open-webui:dev

Xazulula Amaphutha OXhumano lwe-Ollama

Inkinga evamile ye-Docker ivela lapho i-Open WebUI ingakwazi ukufinyelela ku-Ollama ku-127.0.0.1:11434 noma ku-host.docker.internal:11434 ngaphakathi kwe-container. Enye indlela yokusebenza ebhaliwe kumasistimu asekelwayo ukusebenzisa ukuxhumana kwenethiwekhi ye-host:

docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_BASE_URL=http://127.0.0.1:11434 --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Ngalo myalo, finyelela i-Open WebUI ku-http://localhost:8080 esikhundleni se-port 3000.

Uma uxhumano lusahluleka, qinisekisa ukuthi i-Ollama iyasebenza, hlola ikheli layo, bese ubheka imibhalo yokuxazulula izinkinga ye-Open WebUI. Izindawo ze-Docker zingadinga ukucushwa okwengeziwe kwenethiwekhi.

Ukugcinwa Nokuphepha

• Landela umhlahlandlela osemthethweni wokubuyekeza lapho ushintsha noma uthuthukisa ukusatshalaliswa.

• Gcina ivolumu ye-Docker eqhubekayo ixhunyiwe ngesikhathi sokuthuthukisa ukuze ugweme ukulahlekelwa idatha yohlelo lokusebenza.

• Buyekeza ilayisensi yephrojekthi nomlando wamalayisensi ngoba isisekelo sekhodi siqukethe izingxenye ezingaphansi kwamalayisensi amaningi, okuhlanganisa izimfuneko zokuphawula umkhiqizo zekhodi yamanje enelayisensi ye-Open WebUI.

• Bika ubungozi obusolwayo ngokuyimfihlo ngohlelo lwephrojekthi lokudalula ngobuqotho ku-GitHub, hhayi eziteshini zomphakathi zomphakathi.

Isiphetho

I-Open WebUI inikeza isixhumi esibonakalayo esiguquguqukayo sokusebenzisa amamodeli e-AI asekhaya nasesingathwe ngengqalasizinda oyilawulayo. Qala nge-Python 3.11 noma ngokusetshenziswa okuqhubekayo kwe-Docker, xhuma i-Ollama noma i-API ehambisana ne-OpenAI, bese unweba uye ku-RAG, ama-agent, ama-plugin, ukubambisana, ukuhlaziya, kanye nezinsizakalo zokukhiqiza ezikwazi ukukhula. Ukuze uthole ezinye izinketho zokusebenzisa, bheka imibhalo esemthethweni yokuqalisa.