Skip to content

Changelog

June 30, 2026

  • New field extraction from markdown: extract scalar fields from documents you have already fully extracted — no PDF re-upload and no re-processing. It reuses your JSON Schema and the same merge logic as field extraction, then lets you retrieve the values via documents/scalars. See Field extraction from markdown.
  • The Python and Node SDKs now expose extract_fields_from_markdown / extractFieldsFromMarkdown plus documents.scalars for retrieval. See the Python SDK and Node SDK.

June 25, 2026

  • New evaluation: score RAG outputs with Ragas metrics — one-shot batch over a JSONL dataset, plus continuous tracking of a live RAG. See Evaluation.
  • SDKs updated to v0.2.1 — both the Python SDK and Node SDK now expose the evaluation module (one-shot oneshot + continuous tracking).

May 4, 2026

  • SDKs updated to v0.2.0 — both the Python SDK and the new Node SDK now cover the full RAG ingestion pipeline.
  • New Node.js / TypeScript SDK with an async client: reference here.
  • Field-based extraction: extract a structured payload that conforms to a JSON Schema you supply, plus schema generation from a natural-language description.
  • New bucket management: group extracted documents into reusable buckets for the downstream pipeline. See Bucket management.
  • New chunking: split documents into retrieval-sized chunks with section-greedy, markdown-header, or block-window strategies. See Chunking.
  • New embedding: compute dense / sparse vectors per chunk with internal or external models. See Embedding.
  • New vector store: load embeddings into your own Milvus / Zilliz, Qdrant, or Pinecone cluster. See Vector Store.
  • New "turbo" enrichment mode for richer figure and table descriptions.

March 4, 2026

  • New python SDK find the reference here.
  • Zip files of extractions and documents can be downloaded via API too.
  • Possibility to download Zip files from frontend & via API or SDK based on local path and not on firebase storage, for offline use.
  • Filtering results now is possible for a more specific extraction.
  • Now is possible to get a "section summary" instead of a complete "document" summary
  • Added name and description of extractions to identify them better.
  • Improvements on general extraction logic.

October 1, 2025

  • New docs site.
  • Billing & credits guide: how credits, top‑ups, and auto‑recharge work.
  • Quick start tutorial to run your first extraction.
  • API basics to connect your app.

Need help? support@ainexxo.com