Abubuwan iyawa

Abin da kowane model zai iya yi ta API, model bayan model: tools, image input, reasoning, JSON output, adana prompt, binciken yanar gizo da streaming.

Ana isa ga kowane model ta endpoints uku iri ɗaya da maɓalli ɗaya. Teburan biyu suna nuna inda models suka bambanta. Dash yana nufin model yana amsawa ba tare da wannan iyawar ba.

Models na Shannon

Model	Tools	Hotuna	Reasoning	Effort	JSON object	JSON schema	Caching	Bincike	Streaming
`shannon-1.6-lite`	Eh	Eh	—	—	Eh	Eh	—	Eh	Eh
`shannon-1.6-pro`	Eh	Eh	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-2-lite`	Eh	—	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-2-pro`	Eh	—	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-3`	Eh	Eh	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-3-pro`	Eh	Eh	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-3.1`	Eh	Eh	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-3.1-pro`	Eh	Eh	Eh	—	Eh	Eh	—	Eh	Eh
`shannon-3.5`	Eh	—	—	—	Eh	Eh	—	—	Eh
`shannon-coder-1`	Eh	—	Eh	—	Eh	Eh	—	—	Eh

Hosted open-weight models

Id na hosted model shi ne sunansa da bambancinsa da aka haɗa da saƙa (hyphen), misali DeepSeek-V4-Pro-0813-3BIT-REAP.

Model	Tools	Hotuna	Reasoning	Effort	JSON object	JSON schema	Caching	Bincike	Streaming
DeepSeek-V4-Pro-0813 `3BIT-REAP`	Eh	—	Eh	Eh	Eh	Eh	Eh	—	Eh
GLM-5.2 `3BIT-REAP`	Eh	—	Eh	Eh	Eh	Eh	Eh	—	Eh
Kimi-K3 `3BIT-REAP`	Eh	Eh	Eh	Eh	Eh	Eh	Eh	—	Eh
Nemotron3Ultra `3BIT-REAP`	Eh	—	Eh	Eh	Eh	Eh	Eh	—	Eh
MiniMax-M3 `3BIT-REAP`	Eh	Eh	Eh	Eh	Eh	Eh	Eh	—	Eh
DeepSeek-V4-Flash-0731 `W4A16-AUTOROUND-REAP`	Eh	—	Eh	Eh	Eh	Eh	Eh	—	Eh
Kimi-K2.6 `W4A16-AUTOROUND-REAP`	Eh	Eh	Eh	Eh	Eh	Eh	Eh	—	Eh
Laguna-S-2.1 `W4A16-AUTOROUND-REAP`	Eh	—	Eh	Eh	—	—	Eh	—	Eh
inkling `W4A16-AUTOROUND-REAP`	Eh	Eh	Eh	Eh	Eh	—	Eh	—	Eh
MiMo-V2.5-Pro `W8A16`	Eh	—	Eh	Eh	Eh	Eh	Eh	—	Eh
MiMo-V2.5 `W8A16`	Eh	Eh	Eh	Eh	Eh	Eh	Eh	—	Eh
Hy3 `W8A16`	Eh	—	Eh	Eh	Eh	Eh	Eh	—	Eh

Ma'anar kowane ginshiƙi

Tools. Kuna bayyana functions a cikin tools, model yana amsawa da kiran da yake so a yi, kuma code ɗinku yana gudanar da su. Kowane model yana karɓar tools. Hosted open-weight models suna aiwatar da tool_choice na tilas. Kiran Aiki
Image input. Model yana karanta hotuna da aka aika a cikin saƙon mai amfani, a matsayin data URL ko adireshi. shannon-2-lite da shannon-2-pro suna karanta hotuna a requests da kuma suke aika tools ko response_format. Iyalin Shannon 3 yana karanta takardu ma. Hotuna da fayiloli
Reasoning. Amsar na iya ɗaukar rubutun reasoning na model kusa da amsar, a cikin field nasa. Reasoning effort
Reasoning effort. Field na request yana saita zurfin reasoning na model: off, low, medium ko high. Models na Shannon suna saita zurfinsu da kansu. Reasoning effort
JSON object. Da response_format: {"type": "json_object"} amsar JSON ce ba tare da tsari da aka sanya ba. A /v1/chat/completions da /v1/responses. Abubuwan da aka Tsara
JSON schema. Da response_format na nau'in json_schema amsar tana bin schema da kuka aika. A /v1/chat/completions da /v1/responses. Abubuwan da aka Tsara
Adana prompt. Prompt da ya fara kamar na kwanan nan a model iri ɗaya ana karanta shi daga cache, kuma ana cajin ɓangaren da aka yi cache a 25% na farashin input. Ba ya buƙatar field a request. Adana prompt
Binciken yanar gizo. Da web_search: true API yana bincika yanar gizo kuma yana ba model sakamakon kafin ya amsa. A /v1/chat/completions da /v1/messages. Field ne na request kuma ba shi da flag a GET /v1/models. Binciken Yanar Gizon da aka Gina
Streaming. Da stream: true amsar tana zuwa a matsayin server-sent events yayin da model yake rubutu. Kowane model yana yin stream a dukkan endpoints uku. Yawo

Bayanai iri ɗaya daga GET /v1/models

GET https://api.shannon-ai.com/v1/models

GET /v1/models yana lissafa kowane model tare da object ɗin capabilities, don haka shirin kwamfuta zai iya karanta waɗannan bayanai maimakon wannan shafin. Endpoint ɗin ba ya buƙatar maɓalli.

from openai import OpenAI

client = OpenAI(api_key="YOUR_API_KEY", base_url="https://api.shannon-ai.com/v1")

for model in client.models.list():
    print(model.id, getattr(model, "capabilities", None))

import OpenAI from "openai";

const client = new OpenAI({ apiKey: "YOUR_API_KEY", baseURL: "https://api.shannon-ai.com/v1" });

for await (const model of client.models.list()) {
  console.log(model.id, model.capabilities);
}

curl https://api.shannon-ai.com/v1/models

id da capabilities na shigarwa biyu na data:

200 JSON

[
  {
    "id": "shannon-3",
    "capabilities": {
      "tools": true,
      "vision": true,
      "reasoning": true,
      "reasoning_effort": false,
      "streaming": true,
      "json_schema": true
    }
  },
  {
    "id": "Kimi-K3-3BIT-REAP",
    "capabilities": {
      "tools": true,
      "vision": true,
      "reasoning": true,
      "reasoning_effort": true,
      "streaming": true,
      "response_format": true,
      "json_schema": true,
      "prompt_caching": true
    }
  }
]

Ginshiƙi	Flag a capabilities
Tools	`tools`
Image input	`vision`
Reasoning	`reasoning`
Reasoning effort	`reasoning_effort`
JSON object	`response_format`
JSON schema	`json_schema`
Adana prompt	`prompt_caching`
Binciken yanar gizo	Ba flag ba: field ɗin request `web_search`.
Streaming	`streaming`

response_format da prompt_caching flags ne na shigarwar hosted open-weight. Shigarwar Shannon tana bayyana JSON output da json_schema kaɗai, wanda ke wakiltar duka JSON modes.
vision shi ne true ga shannon-2-lite da shannon-2-pro: suna karanta hotuna a requests da kuma suke aika tools ko response_format.

A wajen layin model

Hosted open-weight model yana ƙin hoto ko JSON mode da ke wajen layinsa. Ƙin yana zuwa kafin a cire wani abu daga balance ɗinku:

Matsayi	Nau'i	Saƙo	Lokaci
`400`	`invalid_request_error`	`<id> does not accept image input`	An aika hoto zuwa hosted open-weight model wanda ba shi da image input.
`400`	`invalid_request_error`	`<id> does not accept response_format`	An nemi JSON output daga hosted open-weight model wanda ba shi da kowanne JSON mode.
`400`	`invalid_request_error`	`<id> does not accept response_format json_schema; use json_object`	An aika schema zuwa hosted open-weight model wanda yake da JSON object mode kaɗai.

Model na Shannon yana amsa irin wannan request da matsayi 200: iyawar da ke wajen layinsa ba ta da rawa a amsar. Kuskuren Gudanarwa