Adding Chat history problem

Gemini API Documentation
, ,
1

Hello Gang,

I’m working with Live API (Audio 2 Audio) using gemini-2.5-flash-native-audio-preview-09-2025 model, and I want to adding chat history for staying in context.

For that I tried to store the transcription of the user prompt and also the model response but I don’t know how to inject it in the API session.

Here’s my code if you can help me out

async with client.aio.live.connect(model=MODEL, config=config) as session:

        timings = {}



        turns = history.get_data(user_id=user_id)

        print(turns)



        start = time.perf_counter()

        await session.send_realtime_input(

            audio=types.Blob(data=audio_bytes, mime_type="audio/pcm;rate=16000"),

            contents=turns #Not working

        )

        timings["transcription_time"] = round(time.perf_counter() - start, 4)

        print(f"Sending : {timings['transcription_time']}")



        start = time.perf_counter()

        output_text_response = ""

        input_text = ""

        async for response in session.receive():

            if response.server_content.input_transcription:

                input_text += response.server_content.input_transcription.text

            if response.server_content.output_transcription:

                output_text_response += response.server_content.output_transcription.text

            if response.data is not None:

                wav_chunk = create_wav_chunk(response.data)

                yield wav_chunk

        timings["response_time"] = round(time.perf_counter() - start, 4)

        print(f"Rendering : {timings['response_time']}")



        if user_id not in history.data_store:

            history.data_store[user_id] = []



        history.data_store[user_id].append({"role": "user", "parts": [{"text": input_text}]})

        history.data_store[user_id].append({"role": "model", "parts": [{"text": output_text_response}]})
1 Like
3

Have you found or gotten the solution?