fewertokens

Python SDK for FewerTokens — route OpenAI-compatible chat completions through the gateway to reduce token usage.

Install

pip install fewertokens

From this monorepo (editable):

pip install -e "./fewertokens-sdk[dev]"

Quick start

from fewertokens import FewerTokens

client = FewerTokens(
    api_key="tr_live_...",
    model="gpt-4o",
    llm_api_key="sk-...",
    base_url="http://localhost:8000",  # your FewerTokens API host
)
response = client.chat.completions.create("What is Python?")
print(response.content)
print(response.tokens_saved)

Streaming

for chunk in client.chat.completions.create("Explain Docker", stream=True):
    print(chunk.content, end="", flush=True)

Async

response = await client.chat.completions.acreate("Hello!")

stream = await client.chat.completions.acreate("Hi", stream=True)
async for chunk in stream:
    print(chunk.content, end="")

Specification

See ai-token-reducer/docs/SDK_SPEC.md in the main repo for the full HTTP contract.