Nyumbani / Miongozo / Free LLM API yenye function calling

Free LLM API yenye function calling

Function calling, au tool calling, inaruhusu LLM kuomba msimbo wako uendeshe kazi. Agents, boti za gumzo zenye data ya moja kwa moja na utoaji wa data iliyopangwa, vyote hutumia hili. Katalogi ya FreeLLMAPI ina ofa 190 za model za bure zenye tool calling. Unatuma tools za mtindo wa OpenAI, na lango hulitafsiri kwa kila mtoa huduma.

Msimbo, majina ya modeli na lebo za katalogi ya moja kwa moja zinaonyeshwa kwa Kiingereza.

Tool calling inavyofanya kazi kupitia FreeLLMAPI

  1. Unatuma ujumbe na orodha ya tools. Kila zana ina jina, maelezo na JSON Schema ya parameters zake.
  2. Model hujibu kwa tool_calls na finish_reason: "tool_calls". Haiendeshi chochote yenyewe.
  3. Msimbo wako unaendesha kazi na kuongeza matokeo kama ujumbe wenye role: "tool".
  4. Unatuma mazungumzo tena. Model inatumia matokeo kuandika jibu la mwisho.

Watoa huduma wanaooana na OpenAI, kama Groq, Cerebras, Mistral, OpenRouter na Cloudflare, hupokea ombi kama lilivyo. Kwa Gemini, FreeLLMAPI hutafsiri zana kuwa muundo wa functionDeclarations wa Google’s na kutafsiri jibu kurudi. Msimbo wako ni ule ule kwa kila mtoa huduma.

Baadhi ya model huandika tool call kama maandishi ya kawaida badala ya tool call halisi. FreeLLMAPI huzigundua na kuzibadilisha kuwa tool_calls halisi.

Model za bure zenye tool calling leo

Jedwali linaonyesha model za bure zenye nguvu zaidi zenye tool calling. Data inatoka kwenye katalogi ya moja kwa moja. Kwa agents, angalia pia ukubwa wa muktadha, kwa sababu ufafanuzi wa zana na matokeo hutumia tokens nyingi.

Model za bure zenye tool calling, zenye nguvu zaidi kwanza
ModeliBure kwenyeMuktadhaVikomo vya bure (ofa bora)ZanaUoni
GLM 5.3UnoRouter1M1 maombi/dakika
1 req/min per model
NdiyoHapana
Qwen3.8 MaxACLIDE-€20/month shared internal usage credits · no cardNdiyoHapana
GLM 5.3 FlashNVIDIA NIM, UnoRouter1M40 maombi/dakika
40 RPM
NdiyoNdiyo
Gemini 3.8 FlashGoogle AI Studio1MRecurring provider quotaNdiyoNdiyo
Gemini Flash (latest alias)Google AI Studio-10 maombi/dakika
20 maombi/siku
250,000 tokeni/dakika
~3M
NdiyoNdiyo
Qwen3.8 Flash NextAMD Radeon Cloud262KRecurring daily account allowanceNdiyoHapana
Qwen3.8 FlashElectronHub1M$0.25/week shared recurring creditNdiyoNdiyo
Gemini 3.7 FlashGoogle AI Studio1M10 maombi/dakika
20 maombi/siku
250,000 tokeni/dakika
~3M
NdiyoNdiyo
GPT-5.6 LunaElectronHub1M$0.25/week shared recurring creditNdiyoNdiyo
Dots3 Note PreviewKilo Gateway, OpenRouter512K20 maombi/dakika
50 maombi/siku
50 req/day account-wide (1,000 after a $10 purchase)
NdiyoNdiyo
DeepSeek V4 FlashAMD Radeon Cloud, BazaarLink, ElectronHub1M10 maombi/dakika
50 maombi/siku
Shared rate limit
NdiyoHapana
Qwen3.8 27BElectronHub, Groq, OVH AI Endpoints262K30 maombi/dakika
250 maombi/siku
70,000 tokeni/dakika
~6M
NdiyoNdiyo
Gemini 3.6 FlashGoogle AI Studio1M10 maombi/dakika
20 maombi/siku
250,000 tokeni/dakika
~3M
NdiyoNdiyo
Qwen 3.8 27BCloudflare Workers AI262KShared 10k neurons/dayNdiyoNdiyo
Gemini 3.5 FlashGoogle AI Studio1M10 maombi/dakika
20 maombi/siku
250,000 tokeni/dakika
~3M
NdiyoNdiyo

Inaonyesha 15 kati ya 146 modeli za bure zinazolingana. Tazama modeli zote.

Data ya moja kwa moja kutoka katalogi ya FreeLLMAPI (toleo 2026.11.19, imekaguliwa 2026-11-19). Vikomo vya bure hubadilika mara kwa mara. Katalogi hukaguliwa kila siku.

Hatua ya 1: Sakinisha FreeLLMAPI na uongeze funguo

FreeLLMAPI ni chanzo huria. Inaendeshwa kwenye kompyuta yako kwa Docker. Funguo zako za watoa huduma zinabaki kwenye kompyuta yako.

Sakinisha (macOS au Linux)

curl -fsSL https://freellmapi.co/install.sh | bash

Fungua dashibodi kwenye http://localhost:3001. Ongeza funguo kutoka watoa huduma watatu au zaidi kwenye ukurasa wa Keys. Mzunguko wa zana hutuma maombi kadhaa kwa kazi moja, kwa hivyo funguo zaidi zinatoa nafasi zaidi. Nakili ufunguo wako mmoja kutoka ukurasa uleule.

Hatua ya 2: Endesha mzunguko kamili wa zana

curl: ombi la kwanza

curl http://localhost:3001/v1/chat/completions \
  -H "Authorization: Bearer <unified-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{"role": "user", "content": "What is the weather in Karachi?"}],
    "tools": [{
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "Get the current weather for a city.",
        "parameters": {
          "type": "object",
          "properties": {"city": {"type": "string"}},
          "required": ["city"]
        }
      }
    }]
  }'

Python: rudia hadi model itoe jibu la mwisho

import json
from openai import OpenAI

client = OpenAI(base_url="http://localhost:3001/v1", api_key="<unified-key>")

def get_weather(city: str) -> dict:
    return {"city": city, "temp_c": 32, "condition": "sunny"}  # call a real API here

FUNCS = {"get_weather": get_weather}
tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get the current weather for a city.",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    },
}]
messages = [{"role": "user", "content": "What is the weather in Karachi and Lahore?"}]

for _ in range(5):  # stop after 5 rounds
    msg = client.chat.completions.create(model="auto", messages=messages, tools=tools).choices[0].message
    if not msg.tool_calls:
        print(msg.content)
        break
    messages.append(msg)
    for call in msg.tool_calls:
        try:
            args = json.loads(call.function.arguments or "{}")
            result = FUNCS[call.function.name](**args)
        except Exception as exc:  # bad JSON, unknown tool or wrong arguments
            result = {"error": str(exc)}
        messages.append({"role": "tool", "tool_call_id": call.id, "content": json.dumps(result)})

Tool calling pia inafanya kazi na stream=True. Unapokea vipande vya delta.tool_calls, kisha kipande cha mwisho chenye finish_reason: "tool_calls". Unganisha vipande vya argumenti kabla ya kuchambua JSON.

Tumia zana kupata data iliyopangwa

Tool calling pia ni njia nzuri ya kupata data yenye umbo lisilobadilika. Fafanua zana moja ambayo parameters zake ni sehemu unazotaka. Weka tool_choice kuwa "required", ili model lazima iite zana. Kisha soma argumenti. Hauhitaji kuendesha kazi yoyote.

Python: toa sehemu kwa wito wa zana unaohitajika

import json
from openai import OpenAI

client = OpenAI(base_url="http://localhost:3001/v1", api_key="<unified-key>")
schema = {
    "type": "function",
    "function": {
        "name": "save_contact",
        "description": "Save one contact found in the text.",
        "parameters": {
            "type": "object",
            "properties": {"name": {"type": "string"}, "email": {"type": "string"}},
            "required": ["name", "email"],
        },
    },
}
resp = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Reach Sara Khan at [email protected]."}],
    tools=[schema],
    tool_choice="required",
)
print(json.loads(resp.choices[0].message.tool_calls[0].function.arguments))

Thibitisha argumenti kwa msimbo wako mwenyewe. Model inaweza kuacha sehemu inayohitajika au kutuma aina isiyo sahihi. Ukaguzi ukishindwa, tuma hitilafu kurudi na uulize tena.

Usalama: usiendeshe tool calls zenye athari, kama malipo, barua pepe au kufuta faili, bila ukaguzi. Muulize mtumiaji athibitishe vitendo hivi kwanza.

Mitego

  • Argumenti ni mfuatano: function.arguments ni mfuatano wa JSON.Ichambue, na ushughulikie hitilafu. Model ndogo wakati mwingine hutuma JSON isiyo sahihi.
  • Linganisha vitambulisho: kila ujumbe wa zana unahitaji tool_call_id ya wito wake. Ongeza ujumbe wa assistant wenye tool_calls kabla ya ujumbe wa zana.
  • Wito kadhaa kwa wakati mmoja: model inaweza kurudisha zaidi ya tool call moja kwenye jibu moja. Endesha zote, kisha tuma matokeo yote.
  • Gharama ya tokens: kila ufafanuzi wa zana hutumwa kwenye kila ombi. Zana ishirini kubwa zinaweza kutumia maelfu ya tokens na kufikia kikomo cha chini cha tokens kwa dakika haraka. Tuma zana zinazohitajika kwa kazi pekee.
  • Maelezo wazi: model huchagua zana kwa jina na maelezo yake. Andika sentensi moja fupi inayosema zana inafanya nini na inatumika lini.
  • Simamisha mzunguko: weka idadi ya juu ya raundi. Model dhaifu inaweza kuita zana ile ile tena na tena.
  • Model zisizo na zana: baadhi ya watoa huduma wa bure hawaungi mkono zana. Kwa mfano, AI Horde huondoa zana na kujibu kama gumzo la kawaida.
Model ile ile wakati wa kazi: FreeLLMAPI huweka mazungumzo kwenye model moja kwa dakika 30. Mzunguko wa zana kwa kawaida hubaki kwenye model moja. Kikomo kikilazimisha mabadiliko, model inayofuata hupokea historia kamili ya ujumbe, pamoja na matokeo ya zana.

Utatuzi wa matatizo

TatizoChanzoSuluhisho
Model inajibu kwa maandishi bila kuita zanaModel ni dhaifu kwenye zana, au maelezo si waziBandika model yenye nguvu zaidi kutoka jedwali, au boresha maelezo ya zana
json.JSONDecodeError kwenye argumentiModel ilituma JSON isiyo sahihiShika hitilafu na uitume kurudi kama matokeo ya zana, ili model ijaribu tena
Hitilafu 400 kuhusu ujumbe wa zanatool_call_id hailingani, au ujumbe wa assistant haupoOngeza ujumbe wa assistant kwanza, kisha ujumbe mmoja wa zana kwa kila wito
Hitilafu 429 wakati wa kazi ya agentMzunguko unatuma maombi mengi makubwaOngeza funguo zaidi za watoa huduma, na tuma zana chache
Mzunguko hausimamiModel inaita zana ile ile tenaWeka kikomo cha raundi, na rudisha matokeo wazi
Tumia kila moja ya hizi kwa ufunguo mmoja. FreeLLMAPI ni kipanga njia huria, unachoendesha mwenyewe. Kinaweka tabaka hizi zote za bure nyuma ya kituo kimoja cha OpenAI kinachooana, na kinahamia kwenye nyingine pale kikomo kinapofikiwa. Vinjari katalogi au anza sasa.

Maswali yanayoulizwa mara kwa mara

Ni free LLM API zipi zinaunga mkono function calling?

Gemini, Groq, Cerebras, Mistral, OpenRouter, Cloudflare na watoa huduma wengine wana model za bure zenye tool calling. Katalogi ya FreeLLMAPI ina ofa 190 za model za bure zenye zana leo.

Je, tool calling ni sawa na function calling?

Ndiyo. OpenAI ndiyo iliyotumia kwanza jina function calling. API ya sasa hutumia tools na tool_calls. Majina yote mawili yanamaanisha kwamba model inaomba msimbo wako uendeshe kazi.

Je, tool calling ya Gemini inafanya kazi kwa muundo wa OpenAI?

Ndiyo, kupitia FreeLLMAPI. Lango hutafsiri zana za OpenAI kuwa functionDeclarations za Gemini na kutafsiri jibu kurudi kuwa tool_calls za OpenAI.

Je, naweza kutuma tool calls kwa streaming?

Ndiyo. Kwa stream=True unapokea vipande vya delta.tool_calls na finish_reason ya mwisho ya tool_calls. Unganisha vipande vya argumenti kabla ya kuvichambua.

Miongozo inayohusiana

Vyanzo

Miongozo yote · Tazama modeli zote za bure