Rekoto ya diphetoho

Kopo e nang le web_search: true e reha mehloli ea eona: sources ho /v1/chat/completions le /v1/messages e thathamisa liphetho tseo model e li file, mme ho /v1/chat/completions annotations e tšoaea moo karabo e bolelang e 'ngoe le e 'ngoe teng. Patlo ya webe

web_search: true ho /v1/chat/completions le /v1/messages e etsa web search pele model e araba mme e e fa liphetho e le mehloli e nomoroang. Kopo e fumaneng liphetho e sebelisa patlo e le 'ngoe ea allowance ea hau ea letsatsi le letsatsi. Ha ho e setseng, kopo e arajoa ntle le patlo. Patlo ya webe

Stream e khaoloang ka lehlakoreng la server e lefisoa ka usage e tlalehiloeng ho fihlela sebakeng seo. Karolo e setseng ea reservation e khutlela ho balance ea hau.

shannon-coder-1 e araba ho /v1/responses joalo ka ha e etsa ho li-endpoint tse ling tse peli. Lisebelisoa tsa CLI tsa ho ngola code

Senotlolo sa API se fetotsoeng se emisa ho sebetsa hanghang.

POST /v1/tokenize le POST /v1/messages/count_tokens: palo ea mahala ea tokens bakeng sa li-model tse hosted tsa open-weight. Ho bala tokens

Prompt caching ho li-model tse hosted tsa open-weight. Input e cached e lefisoa ka 25% ea theko ea input mme e tlalehoa ho usage ho li-endpoint tse tharo tsohle: prompt_tokens_details.cached_tokens, input_tokens_details.cached_tokens le cache_read_input_tokens. GET /v1/models e tlaleha pricing.cached_input_per_million_usd le capabilities.prompt_caching bakeng sa li-model tsena. Caching ea Prompt

Li-model tse hosted tsa open-weight: tool_choice e qobelloang (required kapa function e rehiloeng) e hlahisa kopo ea tool. Mekhosi ya mesebetsi

Li-model tse hosted tsa open-weight li amohela molaetsa oa system o seng molaetsa oa pele oa moqoqo.

API e fumaneha e le Tor onion service le ho aterese ea I2P. Likarabo tsa api.shannon-ai.com li nka liaterese tse peli ho li-header tsa Onion-Location le X-I2P-Location. Li-endpoint tse ngotsoeng, Tor le I2P

Li-model tse ncha shannon-3.1 le shannon-3.1-pro, tse thathamisitsoeng ke GET /v1/models. Li-model le litheko

Kopo ea streaming eo client ea eona e itokollang e tsoela pele ho fihlela qetellong mme e baloa. Pele, karabo e ne e emisa ha ho itokolloa.

Likopo ho /v1/responses li hlaha ho usage log.

Li-model tse hosted tsa open-weight li sebelisa li-field tsa sampling: temperature, top_p, seed, stop le max_tokens ho /v1/chat/completions; temperature, top_p le stop_sequences ho /v1/messages; temperature, top_p le max_output_tokens ho /v1/responses. Chat Completions

Usage log e bontša theko ka USD bakeng sa kopo e 'ngoe le e 'ngoe le e le kakaretso.

Boleng ba model bo seng id e phatlalalitsoeng bo arajoa ka 400, mofuta invalid_request_error le molaetsa unknown model: <id> ho li-endpoint tse tharo tsohle. Pele, kopo e joalo e ne e arajoa ke shannon-1.6-lite. Li-model le litheko

/v1 e amohela senotlolo sa API feela. Cookie ea ho kena ea webosaete e ke ke ea baloa moo. Netefatso

Flood protection: likopo tse 120 ka motsotso ka akhaonte ho li-endpoint tse tharo tsa generation. Meeli le balance

Ha model e sa khone ho araba, phoso e nka molaetsa o tsitsitseng The model backend failed to answer. Please retry. Hlahlobo ea balance e sa khoneng ho etsoa e arajoa ka 503 le mofuta api_error. Diphoso

Kopo e se nang streaming e hlolehang e khutlisetsa reservation ea eona ho balance ea hau.

GET /v1/models e tlaleha capabilities.reasoning_effort bakeng sa model e 'ngoe le e 'ngoe. Effort ea reasoning

Li-model tse hosted tsa open-weight: kopo e se nang streaming e sa fumaneng karabo e fela ka 504 le mofuta api_error. Sephetho se se nang letho ke 502. Diphoso

Li-model tse hosted tsa open-weight li amohela input eo model e rehiloeng e e amohelang. Setšoantšo se rometsoeng ho id ea mongolo feela se arajoa ka 400 le <id> does not accept image input. response_format ho id e se nang structured output e arajoa ka 400. GET /v1/models e tlaleha capabilities.vision, response_format le json_schema bakeng sa id e 'ngoe le e 'ngoe. Litšoantšo le lifaele

Likarabo tse se nang streaming li nka mongolo oa reasoning: reasoning_content molaetsa oa /v1/chat/completions, le block ea thinking pele ho mongolo ho /v1/messages. Effort ea reasoning

Li-model tse hosted tsa open-weight li na le litheko tse ncha, ho tloha ho $0.50 ka 1M input tokens le $2.00 ka 1M output tokens. Li-model le litheko

Li-model tse hosted tsa open-weight li na le id e le 'ngoe ka model, li-id tse leshome le metso e 'meli ka kakaretso. Li-model le litheko

Structured output ho shannon-3, shannon-3-pro, shannon-3.1 le shannon-3.1-pro. GET /v1/models e tlaleha capabilities.json_schema: true bakeng sa tsona. Diphumo tse hlophisitsoeng

/v1/responses e bala text.format (json_object kapa json_schema) bakeng sa structured output. Responses API

/v1/responses e khutlisa reasoning ea li-model tsa Shannon: ntho ea output ea reasoning pele ho ntho ea message, le li-event tsa stream response.reasoning_summary_text.delta le response.reasoning_summary_text.done. Responses API

Field ea model ea karabo ke id e phatlalalitsoeng ho endpoint e 'ngoe le e 'ngoe. Ntho ea Responses e nka model le created_at.

GET /v1/models e tlaleha context window ea model e 'ngoe le e 'ngoe. Li-model le litheko

Li-model tse hosted tsa open-weight lia araba ho /v1/messages le /v1/responses hammoho le ho /v1/chat/completions, ka system prompt ea hau le likopo tsa tool. /v1/messages e amohela thinking, mme reasoning_effort ke field ea kopo. Li-model le litheko

Li-body tsa phoso ho /v1 li latela mebopeho ea SDK: ntho ea error e nang le type le message ho li-endpoint tsa sebopeho sa OpenAI, le ntho e tšoanang ka hare ho {"type": "error"} ho /v1/messages. Diphoso

Keep-alive ho li-stream tse bulehileng ka mor'a metsotsoana e 15 e 'ngoe le e 'ngoe: mohala oa maikutlo oa : ping ho /v1/chat/completions, event: ping ho /v1/messages. Phallela

Li-id tsa likopo tsa tool li qala ka call_ ho /v1/chat/completions le /v1/responses, le ka toolu_ ho /v1/messages. Mekhosi ya mesebetsi

/v1/responses: stream ke false ka default. Responses API

max_completion_tokens e amohelwa e le lebitso le leng la max_tokens ho /v1/chat/completions. Chat Completions

Event ea response.completed ea stream ea Responses e nka array ea output. Responses API

Li-entry tsa GET /v1/models li nka display_name, context_window, max_output_tokens, pricing, capabilities le endpoints. Li-model le litheko

Karabo e 'ngoe le e 'ngoe e nka x-request-id, mme boleng boo u bo romelang bo khutla. Li-body tsa likopo li ka ba 32 MiB.

Senotlolo se ka hlakoloa ntle le hore se secha se fanoe. Usage log e arotsoe ka maqephe.

Litšoantšo le litokomane likopong: likarolo tsa image_url ho /v1/chat/completions, li-block tsa image, document le file ho /v1/messages, likarolo tsa input_image ho /v1/responses. Litšoantšo le lifaele

Model e ncha shannon-3-pro. Li-model le litheko

Model e ncha shannon-3. Li-model le litheko

Li-stream tsa /v1/chat/completions li nka reasoning e le delta.reasoning_content. Effort ea reasoning

max_tokens e bolokoa pakeng tsa 1 le 65,536. Boleng bo kantle ho moeli oo bo isoa qetellong e haufi ka ho fetisisa, ha bo hanoe. Chat Completions

shannon-coder-1 e baloa ka likopo ka plan ho /v1/chat/completions le /v1/messages. Ha likopo tsa fensetere li fela, karabo ke 429 e nang le molaetsa Shannon Coder call quota reached. Lisebelisoa tsa CLI tsa ho ngola code

Moqoqo o moholo ho feta context window o arajoa ka 400, mofuta invalid_request_error le code context_length_exceeded. Phoso ka hare ho stream e fihla e le event ea phoso, eseng e le mongolo karabong. Lisebelisoa tsa CLI tsa ho ngola code

Li-endpoint tsa /v1: POST /v1/chat/completions, POST /v1/messages, POST /v1/responses le GET /v1/models. Ke li-endpoint tseo li-SDK tsa OpenAI le Anthropic, Claude Code le Codex CLI li li bitsang. Kakaretso

Likopo tsa tool ka tools ho /v1/chat/completions le /v1/messages, structured output ka response_format, le system e le string kapa e le li-block tsa mongolo ho /v1/messages. Mekhosi ya mesebetsi

Li-id tsa model shannon-1.6-pro, shannon-2-lite, shannon-2-pro le shannon-coder-1, haufi le shannon-1.6-lite. Li-model le litheko

Phatlalatso ea pele ea API: chat ka sebopeho sa OpenAI le sebopeho sa Anthropic, le endpoint ea Responses, e 'ngoe le e 'ngoe e bitsoa ka senotlolo sa API se qalang ka sk-. Netefatso