AI Context Packager — Format Docs for Claude, GPT, Fine-Tuning JSONL
Package multi-document content into Anthropic XML <documents> blocks, OpenAI fine-tuning JSONL datasets, or token-efficient YAML. Optimise context for LLM ingestion in one click. 100% client-side.
Data Privacy
100% Client-Side. Your data never leaves your device. No server transfers. Zero AI training data retention.
Frequently Asked Questions
What is the Anthropic XML format used for?
Anthropic recommends wrapping retrieved documents in <documents><document><source>…</source><document_content>…</document_content></document></documents> blocks when passing multi-document context to Claude. This structure helps the model clearly attribute information to specific sources and reduces hallucination in retrieval-augmented generation (RAG) pipelines.
How does the OpenAI JSONL format work for fine-tuning?
OpenAI fine-tuning expects one JSON object per line, each containing a "messages" array with system, user, and assistant turns. Each document you add becomes a training example where the title becomes the user message and the content becomes the assistant response. Upload the resulting .jsonl file directly to the OpenAI fine-tuning API endpoint.
Why use YAML instead of JSON for LLM prompts?
YAML omits quotes around most strings and uses indentation instead of brackets and braces, making it about 15–20% more token-efficient than equivalent JSON for the same structured content. This matters when operating close to a model's context window limit. The tradeoff is that YAML is slightly harder to generate programmatically, which is why this tool automates the conversion.