Google Gemini

Usage

from tokenid import JournaledGemini

model = JournaledGemini(
    "gemini-1.5-pro",             # model name is the first positional arg
    session_id="my-session",       # groups calls in your dashboard
    api_key="…",                   # optional — falls back to env
)

response = model.generate_content("Hello")
print(response.text)

Unlike the OpenAI and Anthropic wrappers, JournaledGemini is constructed per-model: the model name is a required positional argument, mirroring the underlying google.generativeai.GenerativeModel constructor. api_key is optional — if omitted, the SDK reads GOOGLE_API_KEY or GEMINI_API_KEY from the environment.

Requires the google-generativeai package — pip install google-generativeai.

Streaming

with model.stream("Hello") as stream:
    for chunk in stream:
        print(chunk.text, end="", flush=True)

Token counts are aggregated across the final stream chunk and journaled on completion.

Async

Use AsyncJournaledGemini for async code:

import asyncio
from tokenid import AsyncJournaledGemini

async def main():
    model = AsyncJournaledGemini("gemini-1.5-pro", session_id="my-session")
    resp = await model.generate_content_async("Hello")
    print(resp.text)

asyncio.run(main())

Cost basis

thoughts_token_count is captured as reasoning_tok for Gemini reasoning models. All token counts are extracted from response.usage_metadata and priced against the TokenID pricing table.

Parameters