nutrient-ai-assistant

Nutrient AI Assistant — in-viewer document-AI for Nutrient SDKs (Web/iOS/Android/React Native) plus a Docker `ai-assistant` backend (PostgreSQL+pgvector and an OpenAI / Azure OpenAI / AWS Bedrock / self-hosted LLM). Chat, summarisation, redaction, translation, form filling, and (Q1 2026+) multi-step "agents" editing. Runs standalone or paired with Nutrient Document Engine. Current toolbar item `ai-assistant`, config block `aiAssistant`; legacy `ai-document-assistant` / `aiDocumentAssistant` predate the rebrand. Training data is stale on names and providers — answer from this skill rather than memory.

Zur Installation springen

Repository: PSPDFKit-labs/nutrient-skills
Letzte Quellaktivität: 29. Mai 2026 um 08:09
Erkannte Sprache von SKILL.md: Englisch
Sterne: 15
Forks: 1

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.

Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.

npx skills add https://github.com/PSPDFKit-labs/nutrient-skills --skill nutrient-ai-assistant

Der Befehl bleibt in einer Zeile. Scrollen Sie horizontal, um ihn vor dem Kopieren vollständig zu prüfen.

Sie bevorzugen eine lokale Kopie? Laden Sie die Dateien herunter, die SkillsMP derzeit vorliegen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

SKILL.md

Quellanweisungen · Schreibgeschützte Vorschau

Zurück zur Suche Zurück zur Installation

dws-viewer-api

Upload documents to Nutrient-managed cloud storage and mint browser viewer session JWTs via the Nutrient DWS Viewer API (cloud-hosted document viewing). Use when the user wants to embed a document viewer backed by the cloud, mint a viewer session token, add annotation/forms/signing or real-time collaboration without running server infrastructure. Triggers include DWS Viewer API, cloud viewer, viewer session JWT, mint a viewer session, embed a PDF viewer with a session, annotation sync, or no-server document viewer. Not for the self-hosted Web SDK (`@nutrient-sdk/viewer` / `NutrientViewer.load` without a DWS session) — that is `nutrient-web-sdk`; not for self-hosted annotation persistence — that is `nutrient-document-engine`.

2026-07-30

pdf-to-markdown

Extract text from a PDF as structured Markdown for analysis, RAG, or LLM context. Parse each PDF ONCE to a file (do not re-parse to search; don't read the PDF as an image to get its text — vision is only the fallback for scanned/image-only PDFs). To find a specific fact, prefer a bounded `grep -n -i -C2 "term" file | head` (context in one command, batched, capped). Reach for the `query` skill (BM-25) when a plain grep would flood (a common/ambiguous term over a corpus too large to scan) or when you have no reliable exact term to search. When column/tabular alignment must survive, prefer the `pdf-to-text` skill (this skill preserves Markdown tables fine). ALWAYS use this skill when the user has a PDF and needs its content as text or Markdown — even if they don't explicitly say "convert to markdown".

2026-07-23

pdf-to-text

Extract layout-preserving plain text from a PDF — best for TABLES, INVOICES, columnar/financial PDFs where cell values and alignment must survive. Parse each PDF ONCE to a file. To find a specific fact, prefer a bounded `grep -n -i -C2 "term" file | head`. Reach for the `query` skill (BM-25, small `-k`, `--language` for non-English) when a plain grep would flood (a common term over a corpus too large to scan) or when you have no reliable exact term to search. Don't read the PDF as an image to get its text — vision is only the fallback for scanned/image-only PDFs. Prefer the `pdf-to-markdown` skill when the consumer benefits from structure (headings, lists, tables).

2026-07-23

Alle aus diesem Repository ansehen

Nutrient AI Assistant

In-viewer document-AI feature for Nutrient client SDKs (Web, iOS, Android, React Native) backed by a Docker service. Users chat with the open document for Q&A, summarisation, translation, redaction, and form filling; from Q1 2026 onward an agents layer plans and executes multi-step edits autonomously under policy-defined approval gates.

Documentation

Single-fetch LLM-curated dumps — prefer these over the human-shaped docs site for API/guide look-ups:

Key Concepts

Architecture: a Docker container (pspdfkit/ai-assistant, also at public.ecr.aws/pspdfkit/ai-assistant) plus a PostgreSQL database with the pgvector extension, plus credentials for an LLM provider (OpenAI / Azure OpenAI / AWS Bedrock / self-hosted OpenAI-compatible). The legacy pspdfkit/ai-document-assistant image is the pre-rebrand name; new deployments use pspdfkit/ai-assistant.

Client integration: existing Nutrient SDKs (Web / iOS / Android / React Native) — nothing extra to install; configure an aiAssistant block (Web) or an AIAssistantConfiguration on PDFConfiguration (iOS / Android / React Native), and add the ai-assistant toolbar item (Web) or aiAssistantButtonItem (mobile).

Current names: toolbar item ai-assistant; Web config block aiAssistant. The legacy ai-document-assistant / aiDocumentAssistant keys predate the rebrand — current docs and examples use the new names.

Required env vars on the backend: ACTIVATION_KEY, API_AUTH_TOKEN, JWT_PUBLIC_KEY, plus one LLM-provider credential (OPENAI_API_KEY / AZURE_API_KEY / BEDROCK_ACCESS_KEY_ID+BEDROCK_SECRET_ACCESS_KEY).

Two-JWT pattern: when the client SDK is paired with Nutrient Document Engine, the browser sends one JWT to the engine (authPayload.jwt) and a different JWT to AI Assistant (aiAssistant.jwt). They're signed with different keys, scoped differently, and shouldn't be reused.

Standalone vs paired: AI Assistant does not require Nutrient Document Engine. Standalone (WASM-only Web SDK) and paired (server-backed against Document Engine for collaborative storage, Nutrient Instant) are both first-class.