Condense Tokens

Condense long text sequences of natural language tokens into soft tokens on Subnet 47.

POSTv1/47/organic

Condense text

Condense long text sequences of natural language tokens into soft tokens.

Required attributes

Name
messages
Type
array
Description
Array of message objects with role (user, assistant, or system) and content.

Optional attributes

Name
tier
Type
string
Description
Inference tier. Options: inference_0, inference_1, research. Defaults to research.
Name
target_model
Type
string
Description
Model to use for condensing text. Currently supports mistralai/Mistral-7B-Instruct-v0.2. Defaults to mistralai/Mistral-7B-Instruct-v0.2.
Name
miner_uid
Type
integer
Description
Miner UID. Defaults to -1.
Name
top_incentive
Type
number
Description
Top incentive value. Defaults to 0.9.

This endpoint is authenticated. Get your API key

Request

POST

v1/47/organic

curl -X POST https://api.bitmind.ai/oracle/v1/47/organic \
  -H "Authorization: Bearer {token}" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {
        "role": "user",
        "content": "Tell the user that they need to supply a proper message body"
      }
    ],
    "tier": "research",
    "target_model": "mistralai/Mistral-7B-Instruct-v0.2",
    "miner_uid": -1,
    "top_incentive": 0.9
  }'

Response

{
  "compressed_kv_url": "http://132.145.195.244:10001/1739377231932785714.npy"
}

The condense tokens model

The condense tokens model contains the compressed KV URL.

Properties

Name
compressed_kv_url
Type
string
Description
The URL of the compressed KV.