Uwezo

Kila model hufikiwa kupitia endpoint zile zile tatu kwa key ile ile. Majedwali mawili yanaonyesha mahali model zinapotofautiana. Kistari kinamaanisha model hujibu bila uwezo huo.

Model za Shannon

Model za open-weight zinazopangishwa

Id ya model inayopangishwa ni jina lake na toleo lake vinavyounganishwa kwa kistari, kwa mfano DeepSeek-V4-Pro-0813-3BIT-REAP.

Kila safu inamaanisha nini

• Tools. Unaeleza functions kwenye tools, model hujibu na wito inaotaka ufanywe, na code yako inaziendesha. Kila model inakubali tools. tool_choice ya kulazimisha inatumiwa na model za open-weight zinazopangishwa. Uitoaji wa kazi
• Input ya picha. Model husoma picha zinazotumwa ndani ya ujumbe wa user, kama data URL au kama anwani. shannon-2-lite na shannon-2-pro husoma picha kwenye maombi yanayotuma pia tools au response_format. Familia ya Shannon 3 husoma pia hati. Picha na mafaili
• Reasoning. Jibu linaweza kubeba maandishi ya reasoning ya model kando ya jibu, kwenye field yake yenyewe. Reasoning effort
• Reasoning effort. Field ya ombi huweka kina cha model kufikiri: off, low, medium au high. Model za Shannon huweka kina chao wenyewe. Reasoning effort
• Kitu cha JSON. Kwa response_format: {"type": "json_object"} jibu ni JSON bila umbo lililoagizwa. Kwenye /v1/chat/completions na /v1/responses. Matokeo yaliyopangwa
• JSON schema. Kwa response_format ya aina json_schema jibu hufuata schema unayotuma. Kwenye /v1/chat/completions na /v1/responses. Matokeo yaliyopangwa
• Prompt caching. Prompt inayoanza kama ya hivi karibuni kwenye model ile ile husomwa kutoka cache, na sehemu iliyokeshwa hutozwa kwa 25% ya bei ya input. Haihitaji field ya ombi. Kukesha kwa prompt
• Web search. Kwa web_search: true API hutafuta kwenye wavuti na kuipa model matokeo kabla ya kujibu. Kwenye /v1/chat/completions na /v1/messages. Ni field ya ombi na haina flag kwenye GET /v1/models. Utafutaji wa wavuti
• Streaming. Kwa stream: true jibu hufika kama server-sent events wakati model inaandika. Kila model hufanya stream kwenye endpoint zote tatu. Kutiririsha

Ukweli ule ule kutoka GET /v1/models

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

GET /v1/models inaorodhesha kila model pamoja na kitu cha capabilities, ili programu iweze kusoma ukweli huu badala ya ukurasa huu. Endpoint haihitaji key.

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 na capabilities za entries mbili za data:

[
  {
    "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
    }
  }
]

• response_format na prompt_caching ni flags za entries za open-weight zinazopangishwa. Entry ya Shannon huripoti output ya JSON kwa json_schema peke yake, ambayo inamaanisha hali zote mbili za JSON.
• vision ni true kwa shannon-2-lite na shannon-2-pro: zinasoma picha kwenye maombi yanayotuma pia tools au response_format.

Nje ya safu ya model

Model ya open-weight inayopangishwa hukataa picha au hali ya JSON iliyo nje ya safu yake. Kukataa hutokea kabla ya kitu chochote kutolewa kwenye salio lako:

Model ya Shannon hujibu ombi kama hilo kwa hali 200: uwezo ulio nje ya safu yake hautumiki kwenye jibu. Ushughulikiaji wa Makosa