FastGPT is a Kagi service using powerful LLMs to answer user queries running a full search engine underneath. Think ChatGPT, but on steroids and faster! You can try the web app here.
- Get the API key (requires a Kagi account)
- Top off your API credits.
- Call the API. See example usage.
To obtain an API key you will first need to create a Kagi account.
Then, navigate to Settings -> Advanced -> API portal, or click here to go directly.
Click "Generate API Token".
The FastGPT API uses pre-paid API credits. You can manage API credits in the API billing page of your Kagi settings.
Using API with insufficient credits will produce "Insufficient credit to perform this request." response message.
Pricing for FastGPT is a flat rate per-query:
- 1.5¢ per query ($15 USD per 1000 queries) when
web_search
is enabled. - 0.15¢ per query ($1.5 USD per 1000 queries) when
web_search
is disabled.
Cached responses, as with all other Kagi APIs, are free.
You can reach out through support@kagi.com, our Discord server or KagiFeedback.org.
Answer a query.
Returns a FastGPT Answer.
POST request with web search
curl -v \
-H "Authorization: Bot $TOKEN" \
-H "Content-Type: application/json" \
--data '{"query": "Python 3.11"}' \
https://kagi.com/api/v0/fastgpt
Alternatively in python:
import requests
base_url = 'https://kagi.com/api/v0/fastgpt'
data = {
"query": "python 3.11",
}
headers = {'Authorization': f'Bot {TOKEN}'}
response = requests.post(base_url, headers=headers, json=data)
print(response.json())
Response
{
"meta": {
"id": "120145af-f057-466d-9e6d-7829ac902adc",
"node": "us-east",
"ms": 7943
},
"data": {
"output": "Python 3.11 was released in 2021 and introduced several new features compared to previous versions:
- Faster performance: The interpreter was optimized, resulting in up to 1.25x speedup on benchmarks.[1]
- New syntax features: Exception groups were added using PEP 654.[1]
- Traceback annotation: The new version includes a feature to annotate tracebacks to make them more intelligible.[2]
- Type hint improvements: Type hints were made more powerful.[3]",
"tokens": 757,
"references": [
{"title": "What's New In Python 3.11 — Python 3.11.3 documentation", "snippet": "...", "url": "https://docs.python.org/3/whatsnew/3.11.html"},
{"title": "Introducing the New Features in Python 3.11 -", "snippet": "...", "url": "https://earthly.dev/blog/python-3.11-new-features/"}
]
}
}
Field | Type | Required | Description |
---|---|---|---|
query | string | Yes | A query to be answered. |
web_search | bool | No | Whether to use web search results in answering the query. (default is true) |
cache | bool | No | Whether to allow cached requests & responses. (default is true) |
Field | Type | Description |
---|---|---|
output | string | Answer output |
references | array of Reference object | The search results that are referred in the answer. |
tokens | int | Amount of tokens processed |
Field | Type | Description |
---|---|---|
title | string | Title of the referenced search result. |
snippet | string | Snippet of the referenced search result. |
url | string | URL of the referenced search result. |