Iyakoki da balance

Wannan shafin yana bayanin abin da balance ɗinku ya ƙunsa, abin da request ɗaya ke ware da kashewa, requests nawa za ku iya aikawa, da ƴan iyakokin da request ɗaya zai iya haɗuwa da su.

ƙimar tokens 1M na balance

$5.00

allowance na yau da kullun yana sabuntawa

00:00 UTC

flood protection, ga kowane asusu

120 requests / minti

Yadda ake yi wa requests hidima

• Babu matakan rate — Ƙa'ida ɗaya tana iyakance saurin zuwan requests, kuma iri ɗaya ce ga kowane asusu da kowane plan: requests 120 a minti ɗaya. Babu iyaka kan tokens a minti ɗaya.
• Babu quota na API daban — API yana kashe balance iri ɗaya da na taɗi. Plan yana saita girman allowance na yau. Ba ya saita yawan requests.
• Da sauri yadda kuke so — Ana karɓar requests da aka aika a lokaci guda kuma suna jira a layi. Ba a ƙi su don suna lokaci guda ba.

Balance ɗinku

Ana ƙirga balance ɗinku a tokens. Tokens 1,000,000 na balance sun kai $5.00, kuma kowane farashi a shafin Models & farashi rate ne a kan wannan ƙima.

A kowane lokaci balance shi ne jimillar sassa biyu.

• Plan allowance na yau — Adadin tokens da plan ɗinku ya saita. Yana sabunta kowace rana da ƙarfe 00:00 UTC. Abin da ya rage a ƙarshen rana ba ya ci gaba.
• Credit da aka saya — Tokens da kuka saya a matsayin pack. Credit ba ya ƙarewa kuma yana aiki a kowane plan, har da Free.

• Tsarin kashewa — Kowace request tana kashe plan allowance na yau da farko. Ana amfani da credit da aka saya ne kawai ga abin da ya wuce allowance a wannan rana.
• Taɗi da API suna raba shi — Akwai balance ɗaya ga kowane asusu. Maɓallin API yana kashe daga balance na asusun da ya mallake shi, a farashi iri ɗaya da na taɗi.
• Packs — Ana sayar da credit a packs na tokens 1,000,000 ($5.00), 2,000,000 ($10.00) da 5,000,000 ($25.00), ko a adadin da kuka zaɓa daga tokens 1,000,000 zuwa 100,000,000 a $5.00 kowace 1,000,000.

Karin credit Canza plan

Abin da request ke ware da abin da yake kashewa

• Ware — Idan request ya iso yana ware output budget ɗinsa daga balance ɗinku: max_tokens a /v1/chat/completions da /v1/messages, max_output_tokens a /v1/responses. /v1/chat/completions yana kuma karanta max_completion_tokens. Na asali shi ne 4,096 kuma kewayon shi ne 1 zuwa 65,536.
• Karɓa — Ana karɓar request ne kawai idan abin da aka ware ya shiga cikin abin da ya rage a balance ɗinku. Balance da ya fi sifili amma ya ƙasa da output budget yana samun amsar Quota exceeded. Ku aika max_tokens mafi ƙanƙanta don amfani da sauran.
• Daidaitawa — Idan amsar ta cika, ana maye gurbin abin da aka ware da ainihin caji. Cajin na iya zama ƙasa ko sama da abin da aka ware.
• Mayarwa — Request da ya ƙare da matsayin kuskure yana mayar da abin da aka ware gaba ɗaya.

Ainihin cajin ya dogara da iyalin model.

Ana ɗaukar adadin a USD daga balance ɗinku a tokens a $5.00 kowace 1,000,000, an zagaye zuwa cikakken token.

Ƙirgan tokens da POST /v1/tokenize ko POST /v1/messages/count_tokens kyauta ne kuma ba ya ware komai. Ƙidayar tokens

Inda za a ga balance da amfani

Shafin Maɓallai & amfani yana nuna abin da za ku iya kashewa yanzu, plan allowance na yau, credit ɗinku da aka saya da kashewar API na kwanaki 30 da suka wuce. A ƙasa yana lissafa kowace request da maɓallinku ya yi: lokaci, endpoint, model, cached input, tokens da aka caje da kuɗi. Maɓallai & amfani

Kowace amsa tana kuma ɗauke da object ɗin usage mai ƙirgan tokens na wannan kira.

• usage yana ɗauke da ƙirgan tokens na model. Adadin da aka ɗauka daga balance ɗinku ba ya cikin amsar: shi ne ginshiƙin Tokens da aka caje na jerin requests a Maɓallai & amfani.
• A /v1/messages da hosted open-weight model, input_tokens shi ne ɓangaren input da ba a yi cache ba, cache_read_input_tokens shi ne ɓangaren da aka yi cache kuma cache_creation_input_tokens koyaushe 0 ne.
• Stream a /v1/chat/completions yana ɗauke da usage a chunk ɗinsa na ƙarshe kafin [DONE]. Yawo

Idan balance ya ƙare

Request wanda abin da aka ware masa bai shiga cikin balance ɗinku ba ana amsa shi da matsayi 429, nau'in rate_limit_error da saƙon da ke ƙasa. Ba a caje komai ba. Ana aika amsa iri ɗaya idan balance ya fi sifili amma ya ƙasa da output budget na request.

{
  "error": {
    "type": "rate_limit_error",
    "message": "Quota exceeded. Upgrade your plan at shannon-ai.com/plan"
  }
}

{
  "type": "error",
  "error": {
    "type": "rate_limit_error",
    "message": "Quota exceeded. Upgrade your plan at shannon-ai.com/plan"
  }
}

A /v1/responses object ɗin error na iya ɗaukar code da param ma, duka biyu null.

Abin da za ku iya yi:

• Ku jira plan allowance na gaba da ƙarfe 00:00 UTC.
• Ku ƙara credit. Ana kashe credit bayan plan allowance kuma ba ya ƙarewa. Karin credit
• Ku koma plan mai allowance na yau da kullun mafi girma. Canza plan
• Ku aika max_tokens mafi ƙanƙanta, idan wani balance ya rage: abin da aka ware sai ya ragu.

Allowance na kiran Shannon Coder

Ana ƙirga shannon-coder-1 a /v1/chat/completions da /v1/messages a kira, ba a tokens ba. Kowane plan yana ɗauke da adadin kira ga kowace taga ta awa 4. Request ɗaya kira ɗaya ne.

• Tagogi suna farawa a 00:00, 04:00, 08:00, 12:00, 16:00, 20:00 UTC. Kiran da suka rage a ƙarshen taga ba sa ci gaba.
• Ana ƙirga kira lokacin da aka karɓi request, kafin model ya amsa. Request da ya gaza bayan haka har yanzu ana ƙirga shi a matsayin kira.
• Waɗannan kira ba sa ware tokens kuma ba sa ɗaukar komai daga balance ɗinku. Jerin requests a Maɓallai & amfani yana nuna adadin tokens ɗinsu da ƙimarsu a farashin da aka lissafa.
• max_tokens na asali na shannon-coder-1 a waɗannan endpoints biyu shi ne 65,536.
• Idan babu sauran kira, amsar ita ce matsayi 429, nau'in rate_limit_error, saƙo Shannon Coder call quota reached. Upgrade your plan at shannon-ai.com/plan.
• A /v1/responses, shannon-coder-1 ba shi da allowance na kira: ana cajin sa a tokens daga balance ɗinku a $8.00 kowace 1M, kamar kowane sauran model.

Flood protection

Asusu na iya aika requests 120 a minti ɗaya. Wannan ita ce iyaka ɗaya tilo ga yawan requests, kuma iri ɗaya ce a kowane plan. Tana nan don dakatar da ambaliya, ba don rage amfani na al'ada ba.

• Minti taga ce tsayayye ta daƙiƙa 60 da ke buɗewa da request ɗinku na farko. Idan ta ƙare, ƙirgan yana farawa daga sifili.
• Ana ƙirga ta asusu, ba ta maɓalli ba kuma ba ta adireshin IP ba. Canza maɓalli ba ya buɗe sabuwar taga.
• Request na 121 a cikin taga ana amsa shi da matsayi 429, nau'in rate_limit_error da saƙon Too many requests. Retry in <N>s. N shi ne adadin daƙiƙu har taga ta ƙare, daga 1 zuwa 60.
• Ana duba flood protection kafin balance. Request da ya ƙi ba ya ware komai kuma ba ya kashe komai.

{
  "error": {
    "type": "rate_limit_error",
    "message": "Too many requests. Retry in 37s."
  }
}

{
  "type": "error",
  "error": {
    "type": "rate_limit_error",
    "message": "Too many requests. Retry in 37s."
  }
}

Requests a lokaci guda

Babu iyaka kan requests nawa asusu zai iya buɗewa a lokaci guda, kuma babu kuskure don aika requests a lokaci guda. Requests da ba za su iya farawa nan take ba suna jira a layi kuma ana amsa su bi da bi.

• Kowace request tana ƙirgawa cikin 120 a minti ɗaya idan ta iso, ko da requests na baya sun ƙare ko ba su ƙare ba.
• Kowace request tana riƙe abin da ta ware har ta ƙare. Buɗaɗɗun requests ashirin da output budget na asali suna riƙe 20 × 4,096 = tokens 81,920 na balance. Idan abubuwan da aka ware a haɗe sun fi balance ɗinku girma, request na gaba yana samun amsar Quota exceeded, ko da kiran da suka ƙare za su kashe ƙasa. max_tokens mafi ƙanƙanta yana riƙe ƙasa.
• Request ba tare da streaming ba ba ya aika komai har sai amsarsa ta cika, don haka ku ba client ɗinku timeout da ya rufe jira. Stream yana riƙe haɗinsa a buɗe yayin da yake jira. Yawo

Iyakokin request ɗaya

Kurakurai

Amsoshin wannan shafin. A /v1/messages ana nannaɗe object ɗin error iri ɗaya a matsayin {"type": "error", "error": {…}}.