CODEWITHSUNDEEP
amazon-web-servicesamazon-bedrock
Faruk Kadir
Faruk Kadir

Reputation: 59

Why do I get 'malformed input request' when invoking a Bedrock model?

I'm trying to invoke Bedrock via curl, using this command:

curl --location --request POST 'https://bedrock-runtime.us-east-1.amazonaws.com/model/cohere.command-text-v14/invoke' \
--header 'Content-Type: application/json' \
--header 'Accept: application/json' \
--header 'X-Amz-Content-Sha256: op' \
--header 'X-Amz-Date: 20231104T081154Z' \
--header 'Authorization: 0-HMAC-SHA256 Credential=/20231104/us-east-1/bedrock/aws4_request, SignedHeaders=accept;content-type;host;x-amz-content-sha256;x-amz-date, Signature=' \
--data-raw '{
  "inputText": "Hi",
  "textGenerationConfig": {
    "maxTokenCount": 4096,
    "stopSequences": [],
    "temperature": 0,
    "topP": 0.9
  }
}'

I get the below error:

Malformed input request: 3 schema violations found, please reformat your input and try again.

What is the issue?

Upvotes: 5

Views: 11037

Answers (2)

Kong
Kong

Reputation: 9576

I had exactly this issue and finally managed to get it working by tweaking some rather cryptic parameters.

A better solution, the one I settled on and is recommended by AWS is to use the ConverseCommand API rather than the InvokeModelCommand API. The Converse API standardises the parameters between models.

Upvotes: 1

Ermiya Eskandary
Ermiya Eskandary

Reputation: 23682

You can't use Amazon Titan Model's input for the Cohere Command Model, hence the schema error.

Your request body is for Titan. You can tell this by the inputText field which is unique to this model's inference input. However, you've specified cohere.command-text-v14 as your model ID i.e. Cohere.

Referring to the official docs, the below will work for the Cohere model, with the equivalent parameters set from your Titan input.

curl --location --request POST 'https://bedrock-runtime.us-east-1.amazonaws.com/model/cohere.command-text-v14/invoke' \
--header 'Content-Type: application/json' \
--header 'Accept: application/json' \
--user "$AWS_ACCESS_KEY_ID":"$AWS_SECRET_ACCESS_KEY" \
--aws-sigv4 "aws:amz:us-east-1:bedrock" \
--data-raw '{
  "prompt": "Hi",
  "temperature": 0,
  "p": 0.9,
  "max_tokens": 4096
}'

Output:

{
  "generations": [
    {
      "finish_reason": "COMPLETE",
      "id": "95e12e0f-aa3e-4812-8995-34d124b7354e",
      "text": " Hello! How can I help you today?"
    }
  ],
  "id": "e7fb4f46-a046-4fd5-9bc3-340eefc9007f",
  "prompt": "Hi"
}

Upvotes: 2

Related Questions