Fine-Tuning AI Models on Your Codebase with Vincony

Why Fine-Tune?

General-purpose models don't know your coding conventions, architecture patterns, or domain-specific APIs. A fine-tuned model learns from your codebase and produces suggestions that match your team's style — correct import paths, proper error handling patterns, and accurate internal API usage.

The Fine-Tuning Pipeline

import vincony

client = vincony.Client(api_key="YOUR_KEY")

# Upload training data from your codebase
dataset = client.fine_tuning.upload_dataset(
    source="github",
    repo="your-org/your-repo",
    file_patterns=["*.py", "*.ts"],
    exclude=["tests/*", "node_modules/*"]
)

# Start fine-tuning job
job = client.fine_tuning.create(
    base_model="codestral",
    dataset_id=dataset.id,
    epochs=3,
    learning_rate=1e-5
)

print(f"Job ID: {job.id} — Status: {job.status}")
# Use your custom model: client.chat(model=job.model_id)

What You Can Fine-Tune On

• • Code patterns: Your architecture, naming conventions, patterns
• • Documentation: Internal docs, API references, runbooks
• • Review history: Past PR reviews and approved changes
• • Domain knowledge: Industry-specific terminology and rules

Pricing

Fine-tuning is available on the Business plan ($199/mo). Training costs are based on dataset size and epochs. Inference on your fine-tuned model costs the same as the base model. Contact sales for enterprise volume pricing.