Eventium Event Processing API

Overview

The backend exposes a compact set of endpoints that work together:

• /event — universal ingestion: text, docs, images, audio, and video → structured events.
• /answer — ask questions over those events, with inline citations.
• /spacy — low-level linguistic analysis (tokens, POS, lemmas).

Core Concepts

• Decomposition — each input is broken into clauses (Subject–Verb–Object & metadata).
• Write flag — heuristic indicating whether the input looks like a statement (1) or a question (0).
• Media-aware — images, audio, video, and documents converge to a unified event format.

Endpoints

POST /event

Universal ingestion endpoint. Accepts raw text (text/plain) or multipart/form-data containing text, image, audio, video, or file.

POST /event
Content-Type: multipart/form-data

text: "Extract key events from this document"
file: contract.pdf

POST /answer

Event-based Q&A. Supply the question plus a set of events, and receive a concise answer with citations like [Event 1].

POST /answer
Content-Type: application/json

{
  "question": "What happened in the meeting?",
  "events": [
    "Event 1: The CEO presented the quarterly results.",
    "Event 2: The CFO discussed budget cuts.",
    "Event 3: The team agreed to launch a new product."
  ]
}

POST /spacy

spaCy debug endpoint. Returns tokenized output with POS tags and lemmas.

POST /spacy
Content-Type: text/plain

The contract was signed yesterday.