Audio & Podcast Search Pipeline

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY", namespace="audio-search")

# 1. A bucket for the episodes, and a collection that cuts audio at pauses and transcribes and embeds each segment
bucket = client.buckets.create(
    bucket_name="podcasts",
    bucket_schema={
        "properties": {
            "episode": {
                "type": "audio",
            },
        },
    },
)
collection = client.collections.create(
    collection_name="podcasts",
    source={"type": "bucket", "bucket_ids": [bucket["bucket_id"]]},
    feature_extractor={
        "feature_extractor_name": "multimodal_extractor",
        "version": "v1",
        "parameters": {
            "split_method": "silence",
            "run_transcription": True,
            "run_transcription_embedding": True,
        },
    },
)

# 2. Upload and process
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "episode", "type": "audio", "data": "s3://your-bucket/podcasts/episode-212.mp3"}],
)
client.collections.trigger(collection["collection_id"])

# 3. Semantic search over the transcript embeddings
retriever = client.retrievers.create(
    retriever_name="podcast-search",
    collection_identifiers=["podcasts"],
    input_schema={
        "query": {
            "type": "text",
            "required": True,
        },
    },
    stages=[
        {
            "stage_name": "search",
            "stage_id": "feature_search",
            "parameters": {
                "searches": [
                    {
                        "feature_uri": "mixpeek://multimodal_extractor@v1/multilingual_e5_large_instruct_v1",
                        "query": {
                            "input_mode": "text",
                            "value": "{{INPUT.query}}",
                        },
                        "top_k": 20,
                    },
                ],
                "final_top_k": 20,
            },
        },
    ],
)

# 4. Search
results = client.retrievers.execute(
    retriever["retriever_id"],
    inputs={
    "query": "discussion about AI regulation in Europe",
},
)

for doc in results["documents"]:
    print(doc["start_time"], doc["transcription"], doc["score"])