Search endpoint

The ODEN API has one endpoint. You send a query; you get a synthesized answer and ranked citations.

POST https://api.oden-api.com/search
Authorization: Bearer <your-key>
Content-Type: application/json

Request body#

A concise JSON object. Only query is required.

Field	Type	Default	Description
`query`	string	—	Required. The question or search phrase. Natural language works best. Max 500 characters.
`answer`	boolean	`true`	Include a synthesized answer in `results.answer`. Send `false` for citations only.
`depth`	string	`advanced`	`advanced` does full passage extraction; `basic` returns faster, snippet-level results.
`include_snippets`	boolean	`false`	Add one attributed sentence per citation in `citations[].snippet`.

{
  "query": "latest EU AI Act enforcement timeline",
  "answer": true,
  "depth": "advanced",
  "include_snippets": true
}

Response body#

Field	Type	Description
`query`	string	Your query, echoed back.
`trace_id`	string	Unique diagnostic ID for this request. Include it when reporting an issue.
`cached`	boolean	`true` when the result was served from edge cache in tens of milliseconds.
`results`	object	The answer and its citations.
`results.answer`	string	Synthesized answer, in the same language as the query. Present unless `answer: false`.
`results.citations`	array	Ranked source list, highest relevance first.
`results.citations[].title`	string	Title of the source page.
`results.citations[].url`	string	Canonical URL of the source.
`results.citations[].score`	number	Relevance from 0.0 to 1.0. Higher score denotes stronger semantic proximity.
`results.citations[].snippet`	string	One attributed sentence. Present only when `include_snippets: true`.

{
  "query": "capital of sweden",
  "trace_id": "oden-mquycpae-p3thrk",
  "cached": false,
  "results": {
    "answer": "The capital of Sweden is Stockholm, which is also its largest city.",
    "citations": [
      { "title": "Stockholm — Wikipedia", "url": "https://en.wikipedia.org/wiki/Stockholm", "score": 0.964 }
    ]
  }
}

Choosing search depth#

Choose your depth according to your application latency budget and information density requirements:

Metric / Feature	`depth: "basic"`	`depth: "advanced"`
Typical latency	80 – 180 ms	350 – 650 ms
Content extraction	Snippet-level index scanning	Deep DOM parsing & passage extraction
Synthesis engine	Lightweight answer generation	Full multi-source synthesis
Ideal use case	Autocomplete, routing, entity checks	RAG pipelines, user chatbots, agent tools

advanced (default) reads the candidate pages and extracts the passages that actually answer the query. Use it when the answer quality matters — RAG context, agent tool calls, anything an end user sees.
basic returns snippet-level results faster and cheaper. Use it for autocomplete-style lookups or when you only need to know which sources are relevant.

Query syntax & search operators#

ODEN natively accepts natural language questions while supporting structured precision operators:

Exact phrase matching: Wrap phrases in double quotation marks ("corporate sustainability reporting directive") to enforce exact substring occurrence in sources.
Domain scoping: Use site: prefixes to restrict retrieval to authoritative institutional sources (e.g., site:regeringen.se, site:europa.eu, or site:riksdagen.se).
Domain exclusion: Prefix unwanted domains with a minus sign (e.g., -site:pinterest.com) to suppress social media noise.

{
  "query": "site:europa.eu "AI Act" high-risk classification criteria",
  "depth": "advanced"
}

TypeScript interfaces#

For type-safe integrations, drop this contract directly into your codebase:

export interface OdenSearchRequest {
  query: string;
  answer?: boolean;
  depth?: "basic" | "advanced";
  include_snippets?: boolean;
}

export interface OdenCitation {
  title: string;
  url: string;
  score: number;
  snippet?: string;
}

export interface OdenSearchResponse {
  query: string;
  trace_id: string;
  cached: boolean;
  results: {
    answer?: string;
    citations: OdenCitation[];
  };
}

Notes#

Language follows the query. Ask in Swedish, get a Swedish answer; ask in English, get English, with facts translated from foreign-language sources where needed.
Citations are metadata, not reproduced text. You get titles, URLs and scores — plus one attributed sentence per source if you ask for it. This is what keeps ODEN on the right side of the TDM opt-out.

Next steps#

Work with the output: Answers & citations.
Handle failure: Errors and Rate limits.