Docs / Reference / ask: semantic search

ask: semantic search.

A text or a subject's vector, the model, a floor and a population to the units above the floor.

Figure 1POST /v1/ask/semantic_searchspecified

The caller sends SemanticSearchInput to /v1/ask/semantic_search. The platform returns a card whose body is SemanticSearchAnswer.

POST/v1/ask/semantic_search

Semantic search. The call needs the strategy rung under a token with the ask grant, and it is served by the platform.

The route takes SemanticSearchInput and returns a card whose body is SemanticSearchAnswer.

Request body#

The request body is SemanticSearchInput.

field	type	required
`text`	`text`	no
`subject`	UnitRef	no
`model`	Model	yes
`floor`	`share`	yes
`population`	Predicate	yes

Response#

The route returns a card whose body is SemanticSearchAnswer. Every card ends with a foot that states the basis of each number, the scope of the call, and the time when the facts were true.

Headers#

header	required	meaning
`Scale-Scope`	yes	the scope the call runs in: tenant/, then /brand/, then /strategy/
`Scale-As-True-On`	no	the date the facts must have been true on; defaults to now
`Scale-As-Known-On`	no	the date the facts must have been known on; defaults to now
`Scale-Key`	no	for a change, the key made from the input; a repeat under the same key returns the unit held and writes nothing

A caller needs the strategy rung and a token with the ask grant. The platform serves this route.

Example#

The example sends the smallest body that SemanticSearchInput allows. The build validates it against the schema of SemanticSearchInput.

Example

curl -X POST https://api.scaleintelligence.co/v1/ask/semantic_search \
  -H "Scale-Key: $SCALE_KEY" -H "Scale-Scope: tenant/<id>" \
  -H "content-type: application/json" \
  -d '{ "model": "…", "floor": 0, "population": { "tree": { "form": "group" }, "population": "…", "tree_hash": "…", "register_hash": "…", "compiler_version": "…", "comparator_versions": {} } }'

import { client } from "@scale/sdk";
const answer = await client.ask.semantic_search({
    "model": "…",
    "floor": 0,
    "population": {
      "tree": {
        "form": "group"
      },
      "population": "…",
      "tree_hash": "…",
      "register_hash": "…",
      "compiler_version": "…",
      "comparator_versions": {}
    }
  });
// answer.body is a SemanticSearchAnswer. answer.foot holds the basis, the scope and the time.

import requests
answer = requests.post("https://api.scaleintelligence.co/v1/ask/semantic_search",
    headers={"Scale-Key": KEY, "Scale-Scope": "tenant/<id>"},
    json={
      "model": "…",
      "floor": 0,
      "population": {
        "tree": {
          "form": "group"
        },
        "population": "…",
        "tree_hash": "…",
        "register_hash": "…",
        "compiler_version": "…",
        "comparator_versions": {}
      }
    }).json()

let answer: Card = client.post("https://api.scaleintelligence.co/v1/ask/semantic_search")
    .header("Scale-Key", key).header("Scale-Scope", "tenant/<id>")
    .json(&SemanticSearchInput { /* the fields of the page */ }).send().await?.json().await?;

MCP tool#

The MCP tool si.ask.semantic_search takes the same input and returns the same card. The tool only reads data. An MCP host that renders cards draws this card from ui://scale-intelligence/cards/SemanticSearchAnswer. The tool’s page describes it.

POST /v1/ask/semantic_search
Scale-Scope: …
Scale-As-True-On: …
Scale-As-Known-On: …
Scale-Key: …
content-type: application/json

{
  "model": "…",
  "floor": 0,
  "population": {
    "tree": {
      "form": "group"
    },
    "population": "…",
    "tree_hash": "…",
    "register_hash": "…",
    "compiler_version": "…",
    "comparator_versions": {}
  }
}

Without a key, this page shows the request only. The request body is SemanticSearchInput, and the route returns a card whose body is SemanticSearchAnswer.

Claims#

claim	state	route or tool
The platform serves the route /v1/ask/semantic_search at contract 0476e35e6e275db5.	target	`/v1/ask/semantic_search`