Mbiri ya kusintha

Pempho lokhala ndi web_search: true limatchula magwero ake: sources pa /v1/chat/completions ndi /v1/messages imalemba zotsatira zomwe model idapatsidwa, ndipo pa /v1/chat/completions annotations imasonyeza pomwe yankho limatchula chilichonse. Kusaka pa web

web_search: true pa /v1/chat/completions ndi /v1/messages imasaka pa web model isanayankhe ndipo imapereka zotsatira ngati magwero owerengedwa. Pempho lomwe lapeza zotsatira limagwiritsa ntchito kusaka kumodzi kwa allowance yanu ya tsiku. Ngati palibe chotsalira, pempho limayankhidwa popanda kusaka. Kusaka pa web

Stream yomwe yadulidwa kumbali ya server imaperekedwa mtengo wa usage yomwe yanenedwa mpaka pamenepo. Zotsalira za reservation zimabwerera ku balance yanu.

shannon-coder-1 imayankha pa /v1/responses monga momwe imachitira pa ma endpoint ena awiri. Zida za CLI zolembera code

API key yosinthidwa imasiya kugwira ntchito nthawi yomweyo.

POST /v1/tokenize ndi POST /v1/messages/count_tokens: kuwerenga ma token kwaulere pa hosted open-weight model. Kuwerenga ma token

Prompt caching pa hosted open-weight model. Input yosungidwa imaperekedwa mtengo wa 25% ya mtengo wa input ndipo imanenedwa mu usage pa ma endpoint onse atatu: prompt_tokens_details.cached_tokens, input_tokens_details.cached_tokens ndi cache_read_input_tokens. GET /v1/models imanena pricing.cached_input_per_million_usd ndi capabilities.prompt_caching pa ma model awa. Kusunga prompt

Hosted open-weight model: tool_choice yokakamiza (required kapena function yotchulidwa) imabweretsa call ya tool. Kuyitana mafunkseni

Hosted open-weight model zimalandira message ya system yomwe si message yoyamba ya zokambirana.

API imapezeka ngati Tor onion service komanso pa adilesi ya I2P. Mayankho a api.shannon-ai.com amanyamula ma adilesi onse awiri mu ma header a Onion-Location ndi X-I2P-Location. Ma endpoint obisika, Tor & I2P

Ma model atsopano shannon-3.1 ndi shannon-3.1-pro, olembedwa ndi GET /v1/models. Ma model ndi mitengo

Pempho la stream lomwe client yake yadula limapitirira mpaka kumapeto ndipo limawerengedwa. Kale, yankho linkaima pa kudula.

Ma pempho a /v1/responses amaoneka mu usage log.

Hosted open-weight model zimagwiritsa ntchito ma sampling field: temperature, top_p, seed, stop ndi max_tokens pa /v1/chat/completions; temperature, top_p ndi stop_sequences pa /v1/messages; temperature, top_p ndi max_output_tokens pa /v1/responses. Chat Completions

Usage log imaonetsa mtengo mu USD pa call iliyonse komanso ngati chiwerengero chonse.

Mtengo wa model womwe si id yofalitsidwa umayankhidwa ndi 400, type invalid_request_error ndi message unknown model: <id> pa ma endpoint onse atatu. Kale, pempho loterolo linkayankhidwa ndi shannon-1.6-lite. Ma model ndi mitengo

/v1 imalandira API key yokha. Cookie yolowera ya webusaiti siiwerengedwa kumeneko. Kutsimikizira

Flood protection: ma pempho 120 pa mphindi pa akaunti pa ma endpoint atatu opanga. Malire ndi balance

Model ikalephera kuyankha, cholakwika chimanyamula message yokhazikika The model backend failed to answer. Please retry. Kuyang'ana balance komwe sikungatheke kumayankhidwa ndi 503 ndi type api_error. Zolakwika

Pempho losachita stream lomwe lalephera limabweza reservation yake ku balance yanu.

GET /v1/models imanena capabilities.reasoning_effort pa model iliyonse. Reasoning effort

Hosted open-weight model: pempho losachita stream lomwe silinalandire yankho limatha ndi 504 ndi type api_error. Zotsatira zopanda kanthu ndi 502. Zolakwika

Hosted open-weight model zimalandira input yomwe model yotchulidwa imalandira. Chithunzi chotumizidwa ku id ya text yokha chimayankhidwa ndi 400 ndi <id> does not accept image input. response_format pa id yopanda structured output imayankhidwa ndi 400. GET /v1/models imanena capabilities.vision, response_format ndi json_schema pa id iliyonse. Zithunzi ndi ma file

Mayankho osachita stream amanyamula text ya reasoning: reasoning_content pa message ya /v1/chat/completions, ndi block ya thinking patsogolo pa text pa /v1/messages. Reasoning effort

Hosted open-weight model zili ndi mitengo yatsopano, kuyambira pa $0.50 pa ma token 1M a input ndi $2.00 pa ma token 1M a output. Ma model ndi mitengo

Hosted open-weight model zili ndi id imodzi pa model iliyonse, ma id khumi ndi awiri onse. Ma model ndi mitengo

Structured output pa shannon-3, shannon-3-pro, shannon-3.1 ndi shannon-3.1-pro. GET /v1/models imanena capabilities.json_schema: true pa izo. Zotuluka zokonzedwa

/v1/responses imawerenga text.format (json_object kapena json_schema) pa structured output. Responses API

/v1/responses imabweza reasoning ya ma model a Shannon: item ya output ya reasoning patsogolo pa item ya message, ndi ma event a stream response.reasoning_summary_text.delta ndi response.reasoning_summary_text.done. Responses API

Field ya model ya yankho ndi id yofalitsidwa pa endpoint iliyonse. Object ya Responses imanyamula model ndi created_at.

GET /v1/models imanena context window ya model iliyonse. Ma model ndi mitengo

Hosted open-weight model zimayankha pa /v1/messages ndi /v1/responses komanso pa /v1/chat/completions, ndi system prompt yanu ndi ma call a tool. /v1/messages imalandira thinking, ndipo reasoning_effort ndi field ya pempho. Ma model ndi mitengo

Ma body a zolakwika pa /v1 amatsatira mawonekedwe a SDK: object ya error yokhala ndi type ndi message pa ma endpoint a mtundu wa OpenAI, ndi object yomweyo mkati mwa {"type": "error"} pa /v1/messages. Zolakwika

Keep-alive pa ma stream otseguka masekondi 15 aliwonse: mzere wa comment : ping pa /v1/chat/completions, event: ping pa /v1/messages. Kutsitsa nthawi yeniyeni

Ma id a call a tool amayamba ndi call_ pa /v1/chat/completions ndi /v1/responses, ndi toolu_ pa /v1/messages. Kuyitana mafunkseni

/v1/responses: stream ndi false mwa default. Responses API

max_completion_tokens imalandiridwa ngati dzina lina la max_tokens pa /v1/chat/completions. Chat Completions

Event ya response.completed ya stream ya Responses imanyamula array ya output. Responses API

Ma entry a GET /v1/models amanyamula display_name, context_window, max_output_tokens, pricing, capabilities ndi endpoints. Ma model ndi mitengo

Yankho lililonse limanyamula x-request-id, ndipo mtengo womwe mutumiza umabwezedwa. Ma body a pempho akhoza kufika 32 MiB.

Kiyi ikhoza kuthetsedwa popanda kupatsidwa yatsopano. Usage log imagawidwa m'masamba.

Zithunzi ndi zikalata m'ma pempho: ma part a image_url pa /v1/chat/completions, ma block a image, document ndi file pa /v1/messages, ma part a input_image pa /v1/responses. Zithunzi ndi ma file

Model yatsopano shannon-3-pro. Ma model ndi mitengo

Model yatsopano shannon-3. Ma model ndi mitengo

Ma stream a /v1/chat/completions amanyamula reasoning ngati delta.reasoning_content. Reasoning effort

max_tokens imasungidwa pakati pa 1 ndi 65,536. Mtengo wopitirira malirewo umasunthidwa kumapeto apafupi, osakanidwa. Chat Completions

shannon-coder-1 imawerengedwa m'ma call pa plan pa /v1/chat/completions ndi /v1/messages. Ma call a window akatha, yankho limakhala 429 ndi message Shannon Coder call quota reached. Zida za CLI zolembera code

Zokambirana zazikulu kuposa context window zimayankhidwa ndi 400, type invalid_request_error ndi code context_length_exceeded. Cholakwika mkati mwa stream chimafika ngati event ya error, osati ngati text mu yankho. Zida za CLI zolembera code

Ma endpoint a /v1: POST /v1/chat/completions, POST /v1/messages, POST /v1/responses ndi GET /v1/models. Ndi ma endpoint omwe ma OpenAI ndi Anthropic SDK, Claude Code ndi Codex CLI amaitana. Chidule

Ma call a tool ndi tools pa /v1/chat/completions ndi /v1/messages, structured output ndi response_format, ndi system ngati string kapena ma text block pa /v1/messages. Kuyitana mafunkseni

Ma model id shannon-1.6-pro, shannon-2-lite, shannon-2-pro ndi shannon-coder-1, pambali pa shannon-1.6-lite. Ma model ndi mitengo

Kutulutsidwa koyamba kwa API: chat mu mtundu wa OpenAI ndi mu mtundu wa Anthropic, ndi endpoint ya Responses, iliyonse imaitanidwa ndi API key yoyambira ndi sk-. Kutsimikizira