Context Preparation API

Choose this endpoint when you need a single deterministic pipeline that combines cleaning, safety checking, chunking, relevance ranking, and token-budget-aware packing in one call — rather than stitching together separate tools for each step. It is especially well-suited for RAG pre-processing, agent context assembly, and any workflow where content must be sanitized and contradiction-checked before reaching a language model. Prefer it over generic text splitters when you need safety inspection or hybrid ranking included, and over full LLM calls when you want deterministic, reproducible context preparation without generation.

• Missing required content or documents field returns a validation error
• maxTokens set too low to fit even a single chunk may return empty or truncated content
• Invalid queryEmbedding dimensions cause ranking to fail silently or return unranked chunks
• Extremely large documents exceeding service limits may time out
• Malformed target model object may cause packing to fall back to defaults
• Payment failure or insufficient USDC balance returns a 402 response before processing begins