Text
https://spec.modelcontextprotocol.io/specification/2024-11-05/
Unstrukturierte Daten:
Strukturierte Daten:
Der Text ...
Die Funktion des Textes ...
Die Bilder...
Der Text in den Bildern/OCR
Die Tabellen im Dokument
Die Daten der Diagramme im Dokument
Rekonstruktion von Tabellen
Metadata Extraction (Titel, Autor, Referenzen)
This is a bar chart, which shows the percentage of people in Germany who voted for the populist party in the 2020 and 2024 elections. The x-axis shows the years, while the y-axis shows the percentage of people who voted for the populist party. The bar chart is titled "Durchschmitteln 2020-2024."
The image is a bar chart titled "Median." The chart is divided into three horizontal rows, each representing a month. The x-axis represents the month, while the y-axis represents the values of the data points. The data points are color-coded to represent different values:
Reliable RAG
GraphRAG
Grounding
Semantic Chunking
Faithfulness
Answer Relevance
Context Relevance
User Feedback
Self Critique
Reflection
Agentic RAG
Corrective RAG
Re-ranking
Scoring
Correctness
Reliable RAG
GraphRAG
Grounding
Semantic Chunking
Faithfulness
Answer Relevance
Context Relevance
User Feedback
Self Critique
Reflection
Agentic RAG
Corrective RAG
Re-ranking
Scoring
Backtesting
Chat-Simulation
Correctness
Unstrukturiert | Halb strukturiert | Relational | Agentic |
---|---|---|---|
Dokumente | Metadaten | CRM | Prompt Logging |
Document Summaries | CSV, JSON, Scraped Data | ERP | RAG Logging |
Bild/Audio/Video | Produktdaten/PIM | BI-Daten | Agent Logging |
Scans | Conversations | Order Management | RLHF-Feedback |
Wikis etc | Mails | Knowledge Graphs |
Orchestration
Batching
Inkrementelle Updates
Streaming
Deduplication
Change Data Capture
Parsing / OCR
Splitting / Chunking
Embedding / GraphRAG
Authentifizierung
Monitoring / Alerting
Orchestration
Batching
Inkrementelle Updates
Streaming
Deduplication
Change Data Capture
Parsing / OCR
Splitting / Chunking
Embedding / GraphRAG
Authentifizierung
Monitoring / Alerting
Initialkosten--
Grenzkosten--
Realtime++
Verlässlichkeit++
Compliance & Datenschutz++
Initialkosten--
Grenzkosten--
Realtime++
Verlässlichkeit++
Compliance & Datenschutz++
Orchestration
Batching
Inkrementelle Updates
Streaming
Deduplication
Change Data Capture
Unstrukturiert | Halb strukturiert | Relational | Agentic |
---|---|---|---|
Dokumente | Metadaten | CRM | Prompt Logging |
Document Summaries | CSV, JSON, Scraped Data | ERP | RAG Logging |
Bild/Audio/Video | Produktdaten/PIM | BI-Daten | Agent Logging |
Scans | Conversations | Order Management | RLHF-Feedback |
Wikis etc | Mails | Knowledge Graphs |
Parsing / OCR
Splitting / Chunking
Embedding / GraphRAG
Authentifizierung
Monitoring / Alerting
Vereinfachte Integration verschiedener Modelle
Standardisierte Tool-Nutzung über verschiedene Modelle und Systeme hinweg
Verbesserte Kontextualisierung und Nachvollziehbarkeit
Integration in Telemetrie und Alerting Infrastrukturen
Reduzierte Entwicklungszeit für neue Agenten
// Beispiel eines einfachen MCP-Servers in TypeScript
import { Server } from '@modelcontextprotocol/sdk/server';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio';
const server = new Server(
{
name: 'example-server',
version: '0.1.0',
},
{
capabilities: {
resources: {},
tools: {},
},
}
);
// Tool-Definition
server.setRequestHandler(ListToolsRequestSchema, async () => ({
tools: [
{
name: 'get_weather',
description: 'Get weather forecast for a city',
inputSchema: {
type: 'object',
properties: {
city: { type: 'string' },
days: { type: 'number' },
},
required: ['city'],
},
},
],
}));