GET
/v1/schemas/perplexity/embeddings/v1%2Fembeddings
Accept:
kind inputactivity embeddingshash aad5833f…extracted Nov 20, 2026version currentsource upstream spec ↗
Request body
6 properties · 2 required · 0 $defs| property | type | description | constraints |
|---|---|---|---|
| input* | string | string[] | string[][] | … | Input to embed. pplx-embed-v2 models accept a string, an array of strings, pre-chunked documents (an array of string arrays), or image objects. pplx-embed-v1 models accept only a string or array of strings; a request contains only text or only images. Maximum … | |
| model* | enum | The embedding model to use. Start with pplx-embed-v2-context-0.6b. v1 models will be phased out during 2027. | pplx-embed-v2-context-0.6b | pplx-embed-v2-context-9b | pplx-embed-v1-0.6b | pplx-embed-v1-4b |
| input_type | enum | Required for pplx-embed-v2 models. Use query for search queries and document for content you index. Pre-chunked documents and images require document. | query | document |
| num_chunks | enum | pplx-embed-v2 image inputs only. Splits each image into this many horizontal regions and returns one embedding per region. Small images can return fewer regions. | 1 | 2 | 4 | 8 | 16 | 32 · default: 1 |
| dimensions | integer | Output dimensions. v2: 1024 for pplx-embed-v2-context-0.6b (not MRL-trained); 1024 or 2048 for pplx-embed-v2-context-9b (MRL). v1: multiples of 8 from 128 to 1024 for pplx-embed-v1-0.6b or 2560 for pplx-embed-v1-4b. Defaults to the model's full dimensions. | 128 ≤ n ≤ 2560 |
| encoding_format | enum | Output encoding format for embeddings. base64_int8 returns base64-encoded signed int8 values. base64_binary returns base64-encoded packed binary (1 bit per dimension). | base64_int8 | base64_binary · default: "base64_int8" |
Validate a payload
before you spend tokensshell
$ curl -X POST https://modelschemas.com/v1/validate \
-d '{"provider":"perplexity","endpointId":"v1/embeddings","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}