Batch Processing

Submit and process multiple jobs efficiently using the MicroDC client.

Basic Batch Pattern

from microdc import Client, LLMComplete

client = Client(api_key="mDC_...")

questions = [
    "What is Python?",
    "Explain machine learning",
    "What is a neural network?"
]

# Submit all jobs
job_ids = []
for question in questions:
    job = LLMComplete(model="llama3.3")
    job.set_prompt(question)
    job.metadata = {"question": question}

    job_id = client.send_job(job)
    job_ids.append(job_id)

# Wait for all to complete
client.wait_for_all(timeout=300)

# Collect results
for job_id in job_ids:
    details = client.get_job_details(job_id)
    print(f"Q: {details.metadata['question']}")
    print(f"A: {details.result['choices'][0]['message']['content']}\n")

    if details.is_successful():
        client.acknowledge_job(job_id)

Batch with Callbacks

Process results as they arrive:

from microdc import Client, LLMComplete

completed = []

def on_complete(client: Client, job_id: str):
    details = client.get_job_details(job_id)
    if details.is_successful():
        completed.append({
            "question": details.metadata["question"],
            "answer": details.result['choices'][0]['message']['content']
        })
        client.acknowledge_job(job_id)
        print(f"Completed {len(completed)}/{total} jobs")

client = Client(api_key="mDC_...")
client.set_callback(on_complete)

questions = ["Q1?", "Q2?", "Q3?", "Q4?", "Q5?"]
total = len(questions)

for question in questions:
    job = LLMComplete(model="llama3.3")
    job.set_prompt(question)
    job.metadata = {"question": question}
    client.send_job(job)

client.wait_for_all(timeout=300)
print(f"All done: {len(completed)} results")

Batch Document Processing

from microdc import Client, DocumentCall

client = Client(api_key="mDC_...")

# Upload all files
documents = ["doc1.pdf", "doc2.pdf", "doc3.pdf"]
file_tokens = []

for doc_path in documents:
    upload_result = client.upload_file(doc_path)
    file_tokens.append(upload_result['id'])

# Submit all processing jobs
job_ids = []
for i, token in enumerate(file_tokens):
    job = DocumentCall(model="docling")
    job.add_file(token)
    job.metadata = {"filename": documents[i]}
    job_ids.append(client.send_job(job))

# Wait and collect
client.wait_for_all(timeout=600)

for job_id in job_ids:
    details = client.get_job_details(job_id)
    filename = details.metadata['filename']
    status = "OK" if details.is_successful() else "FAILED"
    print(f"  [{status}] {filename}")
    if details.is_successful():
        client.acknowledge_job(job_id)

Batch Embeddings

from microdc import Client, LLMEmbed

client = Client(api_key="mDC_...")

texts = [f"Document {i} content..." for i in range(100)]
chunk_size = 25

job_ids = []
for i in range(0, len(texts), chunk_size):
    chunk = texts[i:i + chunk_size]
    job = LLMEmbed(model="text-embedding-ada-002")
    job.add_texts(chunk)
    job.metadata = {"start_index": i}
    job_ids.append(client.send_job(job))

client.wait_for_all()

all_embeddings = []
for job_id in job_ids:
    details = client.get_job_details(job_id)
    if details.is_successful():
        all_embeddings.extend(details.result['embeddings'])

print(f"Generated {len(all_embeddings)} embeddings")

Error Handling in Batches

from microdc import Client, LLMComplete

client = Client(api_key="mDC_...")

job_ids = []
for prompt in prompts:
    job = LLMComplete(model="llama3.3")
    job.set_prompt(prompt)
    job_ids.append(client.send_job(job))

client.wait_for_all(timeout=300)

succeeded = []
failed = []
for job_id in job_ids:
    details = client.get_job_details(job_id)
    if details.is_successful():
        succeeded.append(details)
        client.acknowledge_job(job_id)
    else:
        failed.append(details)

print(f"Succeeded: {len(succeeded)}, Failed: {len(failed)}")
for f in failed:
    print(f"  Failed job {f.job_id}: {f.error_message}")

Notebook example

See the Batch Processing notebook for an interactive walkthrough with progress tracking and performance metrics.