Skip to main content
POST
/
compress
/
raw
Compress
curl --request POST \
  --url https://api.example.com/compress/raw/ \
  --header 'Content-Type: application/json' \
  --data '
{
  "context": "<string>",
  "prompt": "<string>",
  "scaledown": {
    "rate": {}
  }
}
'
{
  "compressed_prompt": "<string>",
  "original_prompt_tokens": 123,
  "compressed_prompt_tokens": 123,
  "successful": true,
  "latency_ms": 123,
  "request_metadata": {
    "compression_time_ms": 123,
    "compression_rate": {},
    "prompt_length": 123,
    "compressed_prompt_length": 123
  }
}

​
Overview

The /compress/raw/ endpoint compresses your prompt and context, reducing token count while maintaining the semantic integrity needed for high-quality AI responses. Compression ratios of 50–70% are typical, with no meaningful degradation in downstream model output quality.

​
Request

​
context
string
required
Background information, instructions, or supporting text that provides context for the prompt. This is the content most aggressively compressed — structure and meaning are preserved, but redundancy is removed.
​
prompt
string
required
The main query or question to send to your AI model. Kept intact where possible to preserve intent.
​
scaledown
object
required
Compression configuration.

​
Response

​
compressed_prompt
string
The compressed output, ready to pass directly to your AI model in place of the original context and prompt.
​
original_prompt_tokens
number
Token count of the original input.
​
compressed_prompt_tokens
number
Token count of the compressed output.
​
successful
boolean
Whether the compression completed successfully.
​
latency_ms
number
End-to-end request latency in milliseconds.
​
request_metadata
object

​
Error responses

StatusMeaning
400 Bad RequestMalformed request body or missing required fields.
401 UnauthorizedMissing or invalid x-api-key.
429 Too Many RequestsRate limit exceeded. Back off and retry.
500 Internal Server ErrorCompression service unavailable.

​
Authentication

Include your API key in every request using the x-api-key header. 
-H "x-api-key: <your-api-key>"

​
Examples

​
Auto compression

curl -X POST https://api.scaledown.xyz/compress/raw/ \
  -H "Content-Type: application/json" \
  -H "x-api-key: <your-api-key>" \
  -d '{
    "context": "ScaleDown is a context engineering platform. It compresses AI prompts while preserving semantic integrity...",
    "prompt": "Summarize what ScaleDown does in one sentence.",
    "scaledown": {
      "rate": "auto"
    }
  }'
Response: 
{
  "compressed_prompt": "ScaleDown: context engineering platform, compresses AI prompts, preserves semantic integrity.\n\nSummarize what ScaleDown does in one sentence.",
  "original_prompt_tokens": 150,
  "compressed_prompt_tokens": 65,
  "successful": true,
  "latency_ms": 2341,
  "request_metadata": {
    "compression_time_ms": 2341,
    "compression_rate": "auto",
    "prompt_length": 425,
    "compressed_prompt_length": 189
  }
}

​
Fixed compression rate

Pass a number for rate when you need a guaranteed token budget. 
curl -X POST https://api.scaledown.xyz/compress/raw/ \
  -H "Content-Type: application/json" \
  -H "x-api-key: <your-api-key>" \
  -d '{
    "context": "...",
    "prompt": "What are the key points?",
    "scaledown": {
      "rate": 0.4
    }
  }'

​
Notes

  • "auto" rate is recommended for most use cases. Fixed rates below 0.3 may noticeably affect output quality on dense technical content.
  • The compressed_prompt field is a single string — pass it as the full prompt to your downstream model, replacing both context and prompt.
  • Token counts are estimated using the same tokenizer as the target model family. Exact counts may vary slightly depending on the model you use downstream.