Prepare agent-ready document memory with Knowhere
Use Knowhere to parse complex documents into hierarchy-native, citation-backed memory that agents can retrieve and navigate.
npx skills add agentskillexchange/skills --skill prepare-agent-ready-document-memory-with-knowhere
Use Knowhere when an agent or operator needs to turn messy PDFs, PowerPoint files, office documents, tables, images, and other unstructured sources into persistent document memory for RAG or agentic retrieval. The workflow is to ingest a corpus, route files through the appropriate vision or text parsing track, normalize the results into a shared hierarchy-native schema, preserve source pages and citations, and expose document outlines, structural filters, exact search, fuzzy recall, full reading, assets, and relationships for downstream agents. Invoke this instead of using the product normally when the job is not manual document search, but preparing a reliable evidence substrate that agents can traverse, read, and cite across local or offline document collections. The scope boundary is document parsing plus agent-ready retrieval memory; it is not a generic OCR card, a cloud product listing, a vector database listing, or a broad RAG framework.
What this skill actually does
Inputs and prerequisites: Knowhere Cloud API or self-hosted Knowhere stack, document corpus, configured model or vision model provider for parsing where needed, and an agent/RAG client that consumes the generated memory or MCP-compatible retrieval interface.
Setup notes: Use Knowhere Cloud from https://knowhereto.ai for the managed API, or self-host the Knowhere stack with the upstream Docker Compose instructions linked from the repository. Configure the required model provider, ingest documents, run parsing jobs, then connect agents or RAG clients to the generated hierarchy-native retrieval memory.
Source and verification boundary: use https://docs.knowhereto.ai/ as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a Multi-Framework workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.