Amplify Gen2 LLama integration error

[bogris]

Environment information

appy-contest git:(main) ✗ npx ampx info

System:
  OS: macOS 15.0
  CPU: (8) arm64 Apple M1 Pro
  Memory: 134.86 MB / 16.00 GB
  Shell: /bin/zsh
Binaries:
  Node: 20.10.0 - ~/.nvm/versions/node/v20.10.0/bin/node
  Yarn: 1.22.15 - ~/.npm-packages/bin/yarn
  npm: 10.9.0 - ~/.nvm/versions/node/v20.10.0/bin/npm
  pnpm: 6.11.0 - ~/.npm-packages/bin/pnpm
NPM Packages:
  @aws-amplify/auth-construct: 1.5.0
  @aws-amplify/backend: 1.8.0
  @aws-amplify/backend-auth: 1.4.1
  @aws-amplify/backend-cli: 1.4.2
  @aws-amplify/backend-data: 1.2.1
  @aws-amplify/backend-deployer: 1.1.9
  @aws-amplify/backend-function: 1.8.0
  @aws-amplify/backend-output-schemas: 1.4.0
  @aws-amplify/backend-output-storage: 1.1.3
  @aws-amplify/backend-secret: 1.1.4
  @aws-amplify/backend-storage: 1.2.3
  @aws-amplify/cli-core: 1.2.0
  @aws-amplify/client-config: 1.5.2
  @aws-amplify/deployed-backend-client: 1.4.2
  @aws-amplify/form-generator: 1.0.3
  @aws-amplify/model-generator: 1.0.9
  @aws-amplify/platform-core: 1.2.1
  @aws-amplify/plugin-types: 1.5.0
  @aws-amplify/sandbox: 1.2.6
  @aws-amplify/schema-generator: 1.2.5
  aws-amplify: 6.9.0
  aws-cdk: 2.163.1
  aws-cdk-lib: 2.163.1
  typescript: 5.6.3
AWS environment variables:
  AWS_STS_REGIONAL_ENDPOINTS = regional
  AWS_NODEJS_CONNECTION_REUSE_ENABLED = 1
  AWS_SDK_LOAD_CONFIG = 1
No CDK environment variables
npm notice
npm notice New patch version of npm available! 10.9.0 -> 10.9.1
npm notice Changelog: https://github.com/npm/cli/releases/tag/v10.9.1
npm notice To update run: npm install -g [email protected]
npm notice
npm notice
npm notice New patch version of npm available! 10.9.0 -> 10.9.1
npm notice Changelog: https://github.com/npm/cli/releases/tag/v10.9.1
npm notice To update run: npm install -g [email protected]
npm notice

Describe the bug

with this config for a generation in amplify data:

const aiCheckSong = a
  .generation({
    aiModel: a.ai.model("Llama 3.1 405B Instruct"),
    systemPrompt: `You are a helpful assistant that checks, corrects and returns full song names, full composers name  and types. 
      Based on the input, you should return the corrected song name, composer and type. 
      The song name and type should be included in response only if the input contains a song name, 
      Feel free to add the composer if it's not included in the input and you are confident you know it.`,
  })
  .arguments({
    inputName: a.string(),
    inputComposer: a.string(),
  })
  .returns(
    a.customType({
      songName: a.string(),
      songComposer: a.string(),
    })
  )
  .authorization((allow) => allow.publicApiKey());

call from the frontend into api.generation.aiCheckSong will throw an error:

[
{
  "path": [
    "aiCheckSong"
  ],
  "data": null,
  "errorType": "ValidationException:http://internal.amazon.com/coral/com.amazon.bedrock/",
  "errorInfo": null,
  "locations": [
    {
      "line": 2,
      "column": 3,
      "sourceName": null
    }
  ],
  "message": "A custom error was thrown from a mapping template."
}
]

the same set-up works with Antropic Claude 3.5 as the model, and I get the correct formatted object.

Looking at the docs I don't see any limitation of the Llama model.

Reproduction steps

deploy a data api with the setup from above

from FE call the api with:

   const res = await clientPublicApi.generations.aiCheckSong({
      inputName: name,
      inputComposer: composer,
    });
    ```
    both params are strings. 

[bogris]

on closer inspection, looks like the models typed in Amplify are not available in "eu-central-1" and the models that we do have, are not in the enum used to complete the model.

[ykethan]

Hey,👋 thanks for raising this! I'm going to transfer this over to our API repository for better assistance 🙂

[bogris]

i will add that I made it work with llama 3.2 3b available in eu-central-1 by updating the types and model mapping in the library, but I found that this specific one is only usable trough inference profiles. Seems that this is only configurable in conversations. Is there any plan to add it to generation?

10x and sorry for adding another dimension to this issue :)

[ykethan]

@bogris refer to this example aws-amplify/docs#8121 (comment) providing additional information while the team gets a chance to into this issue.

[bogris]

@ykethan thanks for the update

I went trough this example, but it is applicable only for conversations.

in the case of generations, that parameter is not available. I am on backend 1.8 that is latest on npm

here is the typing for .generation()

export interface GenerationInput {
    aiModel: AiModel;
    systemPrompt: string;
    inferenceConfiguration?: InferenceConfiguration;
}```

[tejas2008]

Hi @bogris, the generation route doesn’t currently support Llama 3.1 405B Instruct, which is causing the error. For more details on supported models and routes, please check this link. Let me know if you have further questions!