Vai al contenuto
TakoAPI
Torna alle skill

gradient-knowledge-base

di simondelorean

Web & Frontend Development
—Non ancora valutato
La tua valutazione

Sintesi

Community skill (unofficial) for DigitalOcean Gradient Knowledge Bases.

Descrizione

Cosa fa questa skill

Builds RAG pipelines on DigitalOcean's Gradient AI platform. Handles document storage in DO Spaces, Knowledge Base management via the DO API, and semantic or hybrid search queries against indexed content. Optionally synthesizes answers through DigitalOcean's LLM inference endpoint.

Combines managed vector indexing, S3-compatible object storage, and LLM synthesis entirely within DigitalOcean, eliminating the need to stitch together separate vector database and storage services.

Quando usarla

  • Indexing internal research reports for natural language search
  • Building a Q&A system over company documentation
  • Running semantic queries on uploaded earnings filings
  • Managing multiple knowledge bases across DO regions
  • Automating document upload and re-indexing pipelines

Esempio di workflow

INPUT

User asks: Summarize the key findings from our Q4 research reports

AGENT
  1. 1

    Uploads Q4 report files to a DO Spaces bucket

  2. 2

    Creates a Knowledge Base and connects the bucket as a data source

  3. 3

    Triggers indexing to make the documents searchable

  4. 4

    Runs a RAG query with --rag flag against the Knowledge Base

  5. 5

    Receives a synthesized answer with document citations from the LLM

OUTPUT

A summarized answer with source citations drawn from the indexed Q4 reports

Requisiti

  • DO_API_TOKEN environment variable
  • DO_SPACES_ACCESS_KEY environment variable
  • DO_SPACES_SECRET_KEY environment variable
  • GRADIENT_API_KEY environment variable (required for RAG queries)
  • DigitalOcean account with Gradient AI and Spaces access

Installazione

clawhub install simondelorean/gradient-knowledge-base

Esegui il comando CLI oppure incolla il prompt in Claude Code, Cursor, Codex o nell'agente di coding che usi.

Link