Ilokhi yezinguquko

Isicelo esine-web_search: true sibiza imithombo yaso: i-sources ku-/v1/chat/completions na-/v1/messages ibala imiphumela i-model eyinikeziwe, futhi ku-/v1/chat/completions i-annotations iphawula lapho impendulo icaphuna khona ngayinye. Usesho lwewebhu

I-web_search: true ku-/v1/chat/completions na-/v1/messages yenza usesho lwewebhu ngaphambi kokuba i-model iphendule bese iyinika imiphumela njengemithombo enezinombolo. Isicelo esithole imiphumela sisebenzisa usesho olulodwa lwe-allowance yakho yansuku zonke. Uma kungasali lutho, isicelo siphendulwa ngaphandle kosesho. Usesho lwewebhu

I-stream enqunyiwe ohlangothini lwe-server ibiziswa ukusetshenziswa okubikiwe kuze kube yilelo phoyinti. Insalela yokubekelwa eceleni ibuyela ku-balance yakho.

I-shannon-coder-1 iphendula ku-/v1/responses njengoba yenza kuma-endpoint amanye amabili. Amathuluzi e-CLI okukhoda

I-API key eshintshiwe iyeka ukusebenza ngokushesha.

POST /v1/tokenize ne-POST /v1/messages/count_tokens: ukubalwa kwama-token mahhala kuma-model e-open-weight ahlinzekiwe. Ukubala ama-token

I-prompt caching kuma-model e-open-weight ahlinzekiwe. Okufakwayo okugcinwe kubiziswa ku-25% yentengo yokufakwayo futhi kubikwa ku-usage kuwo wonke ama-endpoint amathathu: prompt_tokens_details.cached_tokens, input_tokens_details.cached_tokens ne-cache_read_input_tokens. I-GET /v1/models ibika i-pricing.cached_input_per_million_usd ne-capabilities.prompt_caching kulawa ma-model. I-Prompt caching

Ama-model e-open-weight ahlinzekiwe: i-tool_choice ephoqelelwe (required noma umsebenzi onegama) ikhiqiza ukubiza ithuluzi. Ukubiza umsebenzi

Ama-model e-open-weight ahlinzekiwe amukela umlayezo we-system ongeyona owokuqala engxoxweni.

I-API ifinyeleleka njengesevisi ye-onion ye-Tor nakukheli le-I2P. Izimpendulo ze-api.shannon-ai.com ziphatha amakheli amabili kuma-header e-Onion-Location ne-X-I2P-Location. Ama-endpoint abethelwe, e-Tor ne-I2P

Ama-model amasha shannon-3.1 no-shannon-3.1-pro, alandelisiwe yi-GET /v1/models. Ama-model namanani

Isicelo esisakazwayo iklayenti laso elinqamukayo siqhubeka size siphele futhi siyabalwa. Ngaphambilini, impendulo yayiyeka lapho kunqamuka.

Izicelo ku-/v1/responses ziyavela erekhodini lokusetshenziswa.

Ama-model e-open-weight ahlinzekiwe asebenzisa ama-field e-sampling: temperature, top_p, seed, stop ne-max_tokens ku-/v1/chat/completions; temperature, top_p ne-stop_sequences ku-/v1/messages; temperature, top_p ne-max_output_tokens ku-/v1/responses. Chat Completions

Irekhodi lokusetshenziswa libonisa izindleko nge-USD ekubizeni ngakunye ngokuphelele.

Inani le-model elingeyona i-id eshicilelwe liphendulwa nge-400, uhlobo invalid_request_error nomlayezo unknown model: <id> kuwo wonke ama-endpoint amathathu. Ngaphambilini, isicelo esinjalo sasiphendulwa yi-shannon-1.6-lite. Ama-model namanani

I-/v1 yamukela i-API key kuphela. I-cookie yokungena yewebhusayithi ayifundwa lapho. Ukuqinisekisa

I-flood protection: izicelo ezingu-120 ngomzuzu nge-akhawunti kuma-endpoint amathathu okukhiqiza. Imikhawulo ne-balance

Uma i-model ingakwazi ukuphendula, iphutha liphatha umlayezo omisiwe othi The model backend failed to answer. Please retry. Ukuhlola i-balance okungenakwenziwa kuphendulwa nge-503 nohlobo api_error. Ukuphathwa kwamaphutha

Isicelo esingasakazwa esihluleka sibuyisela ukubekelwa eceleni ku-balance yakho.

I-GET /v1/models ibika i-capabilities.reasoning_effort ku-model ngayinye. Umzamo wokucabanga

Ama-model e-open-weight ahlinzekiwe: isicelo esingasakazwa esingathola mpendulo siphela nge-504 nohlobo api_error. Umphumela ongenalutho ngu-502. Ukuphathwa kwamaphutha

Ama-model e-open-weight ahlinzekiwe amukela okufakwayo i-model eqanjwe ngayo ekwamukelayo. Isithombe esithunyelwe ku-id yombhalo kuphela siphendulwa nge-400 ne-<id> does not accept image input. I-response_format ku-id engenakho okuphumayo okuhleliwe iphendulwa nge-400. I-GET /v1/models ibika i-capabilities.vision, response_format ne-json_schema ku-id ngayinye. Izithombe namafayela

Izimpendulo ezingasakazwa ziphatha umbhalo wokucabanga: i-reasoning_content emlayezweni we-/v1/chat/completions, ne-block ye-thinking ngaphambi kombhalo ku-/v1/messages. Umzamo wokucabanga

Ama-model e-open-weight ahlinzekiwe anamanani amasha, kusukela ku-$0.50 ngama-token okufakwayo ayi-1M no-$2.00 ngama-token okuphumayo ayi-1M. Ama-model namanani

Ama-model e-open-weight ahlinzekiwe anayo i-id eyodwa nge-model ngayinye, ama-id ayishumi nambili sezizonke. Ama-model namanani

Okuphumayo okuhleliwe ku-shannon-3, shannon-3-pro, shannon-3.1 na-shannon-3.1-pro. I-GET /v1/models ibika i-capabilities.json_schema: true ngazo. Imiphumela ehlanganisiwe

I-/v1/responses ifunda i-text.format (json_object noma json_schema) ngokuphumayo okuhleliwe. I-Responses API

I-/v1/responses ibuyisela ukucabanga kwama-model e-Shannon: into yokuphumayo ye-reasoning ngaphambi kwento ye-message, nemicimbi ye-stream response.reasoning_summary_text.delta ne-response.reasoning_summary_text.done. I-Responses API

I-field ye-model yempendulo iyi-id eshicilelwe kuwo wonke ama-endpoint. Into ye-Responses iphatha i-model ne-created_at.

I-GET /v1/models ibika iwindi lomongo le-model ngayinye. Ama-model namanani

Ama-model e-open-weight ahlinzekiwe aphendula ku-/v1/messages na-/v1/responses kanye nasekhasini /v1/chat/completions, nge-system prompt yakho nokubiza amathuluzi. I-/v1/messages yamukela i-thinking, futhi i-reasoning_effort iyi-field yesicelo. Ama-model namanani

Imizimba yamaphutha ku-/v1 ilandela izimo ze-SDK: into ye-error ene-type ne-message kuma-endpoint efomethi ye-OpenAI, nento efanayo ngaphakathi kwe-{"type": "error"} ku-/v1/messages. Ukuphathwa kwamaphutha

I-Keep-alive ku-stream evulekile njalo emizuzwana engu-15: umugqa wokuphawula we-: ping ku-/v1/chat/completions, i-event: ping ku-/v1/messages. Ukusakaza

Ama-id okubiza amathuluzi aqala nge-call_ ku-/v1/chat/completions na-/v1/responses, nge-toolu_ ku-/v1/messages. Ukubiza umsebenzi

/v1/responses: i-stream ngokuzenzakalelayo ithi false. I-Responses API

I-max_completion_tokens iyamukelwa njengelinye igama le-max_tokens ku-/v1/chat/completions. Chat Completions

Umcimbi we-response.completed we-stream ye-Responses uphatha uhlu lwe-output. I-Responses API

Okufakiwe kwe-GET /v1/models kuphatha i-display_name, context_window, max_output_tokens, pricing, capabilities ne-endpoints. Ama-model namanani

Impendulo ngayinye iphatha i-x-request-id, futhi inani olithumelayo liyabuyiswa. Imizimba yezicelo ingafika ku-32 MiB.

I-key ingahoxiswa ngaphandle kokukhishwa kwenye entsha. Irekhodi lokusetshenziswa lihlukaniswe ngamakhasi.

Izithombe namadokhumenti ezicelweni: izingxenye ze-image_url ku-/v1/chat/completions, ama-block e-image, document ne-file ku-/v1/messages, izingxenye ze-input_image ku-/v1/responses. Izithombe namafayela

I-model entsha shannon-3-pro. Ama-model namanani

I-model entsha shannon-3. Ama-model namanani

Ama-stream ku-/v1/chat/completions aphatha ukucabanga njenge-delta.reasoning_content. Umzamo wokucabanga

I-max_tokens igcinwa phakathi kuka-1 no-65,536. Inani elingaphandle kwalolo hlobo liletshwa ekugcineni okuseduzane, alinqatshwa. Chat Completions

I-shannon-coder-1 ibalwa ngezicelo ngepulani ku-/v1/chat/completions na-/v1/messages. Uma izicelo zewindi sezisetshenzisiwe, impendulo ithi 429 nomlayezo Shannon Coder call quota reached. Amathuluzi e-CLI okukhoda

Ingxoxo enkulu kunewindi lomongo iphendulwa nge-400, uhlobo invalid_request_error nekhodi context_length_exceeded. Iphutha ngaphakathi kwe-stream lifika njengomcimbi wephutha, hhayi njengombhalo empendulweni. Amathuluzi e-CLI okukhoda

Ama-endpoint e-/v1: POST /v1/chat/completions, POST /v1/messages, POST /v1/responses ne-GET /v1/models. Ayizona ama-endpoint abizwa yi-OpenAI ne-Anthropic SDK, i-Claude Code ne-Codex CLI. Uhlaka

Ukubiza amathuluzi nge-tools ku-/v1/chat/completions na-/v1/messages, okuphumayo okuhleliwe nge-response_format, ne-system njenge-string noma njengama-block ombhalo ku-/v1/messages. Ukubiza umsebenzi

Ama-id e-model shannon-1.6-pro, shannon-2-lite, shannon-2-pro ne-shannon-coder-1, eceleni kwe-shannon-1.6-lite. Ama-model namanani

Ukukhishwa kokuqala kwe-API: ingxoxo ngefomethi ye-OpenAI nangefomethi ye-Anthropic, ne-endpoint ye-Responses, ngayinye ibizwa nge-API key eqala nge-sk-. Ukuqinisekisa