Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

fix(generation): gracefully handle stringified tool_use responses #2919

Merged
merged 10 commits into from
Oct 8, 2024

Conversation

atierian
Copy link
Member

@atierian atierian commented Oct 2, 2024

Description of changes

1. Handling stringified JSON responses

Problem:
Bedrock occasionally returns stringified JSON rather than JSON for tool_use input.

Change:
If the type of the Bedrock tool_use response is string and the JS representation of the GraphQL response type in not a string (String, ID, AWSJSON), parse it before returning.

2. Handling AccessDeniedExceptions

Problem:
Bedrock models must be enabled in the console for each region they're being used in.

Change:
Catch AccessDeniedException error types and return a GraphQL error with a generic, but helpful, error message.

CDK / CloudFormation Parameters Changed

N/A

Issue #, if available

N/A

Description of how you validated changes

Checklist

  • PR description included
  • yarn test passes
  • E2E test run linked
  • Tests are changed or added
  • Relevant documentation is changed or added (and PR referenced)
  • New AWS SDK calls or CloudFormation actions have been added to relevant test and service IAM policies
  • Any CDK or CloudFormation parameter changes are called out explicitly

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

@atierian atierian force-pushed the ai.fix-stringified-tool-use branch 2 times, most recently from 99151e5 to b68a235 Compare October 3, 2024 14:23
@atierian atierian marked this pull request as ready for review October 3, 2024 18:48
@atierian atierian requested a review from a team as a code owner October 3, 2024 18:48
@atierian atierian merged commit db34e42 into main Oct 8, 2024
7 checks passed
@atierian atierian deleted the ai.fix-stringified-tool-use branch October 8, 2024 19:03
palpatim added a commit that referenced this pull request Oct 10, 2024
* chore: update .jsii assembly

* chore: update .jsii assembly

* chore: migrate pg array objects e2e test in gen2 cdk (#2906)

* chore: graphql prep for test migration

* refactor: generic graphql field selection string with fieldmap

* feat: add postgres array objects e2e test

* test: remove bootstrap in test code

* chore: schema cleanup

* chore: final cleanup

* chore: add explanation on FieldMap ans examples

* chore: remove dup test

---------

Signed-off-by: Kevin Shan <siqishan@amazon.com>
Co-authored-by: Tim Schmelter <schmelte@amazon.com>

* fix(model-transformer) IndexName -> index in query list resolver (#2912)

* chore: upgrade cdk library dependency to 2.158.0 (#2876)

* chore: upgrade cdk dependency to 2.158.0

* chore: install and use nvm

* chore: use full version for nvm

* chore: testing linux build with nvm

* chore: fix version in cdk tests

* chore: update jsii files

* update: increase memory size

* add: debug statement

* update: mem size back to 8096, use ps1 file for shell script

* fix: path to Setup-NodeVersion.ps1

* fix: path to codebuild_specs/Setup-NodeVersion.ps1

* add: set runtime version

* update: image

* add: debug statement

* update: use earlier code

* add: debug statements

* update: clean up code

* update: use the correct image

* add: list installed node versions and used nodejs.install

* restart: install nvm using choco

* add: back mem size variable

* add: nvm install and use 18.20.4

* add: env var NVM_HOME and NVM_SYMLINK

* add: spawn powershell as admin

* update: remove all other builds

* add: debug statement

* add: env var path

* update: print env var

* add: commands

* update: env var set up

* add: refresh env var

* update: more debug statement

* update

* revamp: find nvm.exe

* update: install nvm windows directly

* update: launch new shell if current shell does not recognize nvm

* update: install node in buildspec

* add: install and use node in build spec

* update: use single quote to prevent interpreting \

* add: 2 scripts, one for installing nvm, another for using nvm

* fix: path error

* test: which way set env var

* update: set up env var in pre_build

* update: use choco in pre-build

* fix: syntax error

* update: build_windows working, running all tests

* test: remove bootstrap in test code

* debug: _runGqlE2ETests

* update: debug_workflow

* update: debug_workflow

* update: debug_workflow

* update: debug_workflow

* add: debug statement

* add: debug test

* add: debug

* update: use uuid for bucket name

* remove: use of uuid

* add: debug statement

* update: use differrent bucket name

* add: mili second timestamp

* add: debug statement

* remove: debug statement

* remove: redundant code

---------

Co-authored-by: Bobby Yu <bobbyu@amazon.com>
Co-authored-by: Tim Schmelter <schmelte@amazon.com>

* test: fix gen 1 init (#2924)

* fix(conversation): allow changes to systemPrompt, inferenceConfig, aiModel to be hotswapped (#2923)

* feat: auto increment support (#2883)

* chore(graphql-default-value-transformer): tidy tests

* test(graphql-default-value-transformer): add unit tests for auto increment support

* feat: 🎸 utils to detect Postgres datasource

* feat: 🎸 support auto increment

Implements support for auto increment (serial) fields from Postgres
datasources. Such fields are denoted by an empty `@default` applied to
an `Int` field.

* test(graphql-default-value-transformer): pk can be auto increment

* test(graphql-default-value-transformer): auto-increment crud e2e

* chore: describe test purpose

* chore: removing logging

* chore: describe why invalid cases are invalid

* chore: remove unecessary e2e test case

* chore: test messaging clarity

* chore: type safety

* chore: alphabetize list

* chore: type of return value asserts against string

Co-authored-by: Tim Schmelter <schmelte+github@amazon.com>

* chore: test ensures customers can insert to serial fields with custom values

* chore: verify that @default(value) works on mysql

* chore: remove unecessary ssm test case

* chore: update branch from main

* test: value cannot be null on ddb

---------

Co-authored-by: Tim Schmelter <schmelte+github@amazon.com>

* fix(conversation): use functionMap for custom handler IFunction reference (#2922)

* fix(generation): gracefully handle stringified tool_use responses (#2919)

* feat(conversation): per message items and lambda history retrieval pattern (#2914)

* fix: sql default value e2e failures (#2932)

* fix(generation): remove trailing comma in inferenceConfig resolver code (#2933)

* fix: add aws_iam to custom operations when enableIamAuthorization is enabled; fix graphql type utils (#2921)

- test: Add additional tests to fix coverage metrics for unchanged files
- test: Add implicit IAM auth support tests
  - Added a skipped test for custom type support, to be re-enabled once we
    figure out the right strategy for this.

* fix: appsync ttl correct duration time unit in ms (#2928)

Signed-off-by: Kevin Shan <siqishan@amazon.com>

---------

Signed-off-by: Kevin Shan <siqishan@amazon.com>
Co-authored-by: amplify-data-ci <amplify-data-dev+github@amazon.com>
Co-authored-by: Kevin Shan <siqishan@amazon.com>
Co-authored-by: Ian Saultz <52051793+atierian@users.noreply.github.com>
Co-authored-by: Phani Srikar Edupuganti <55896475+phani-srikar@users.noreply.github.com>
Co-authored-by: Bobby Yu <bobbyu@amazon.com>
Co-authored-by: Dane Pilcher <dppilche@amazon.com>
Co-authored-by: Peter V. <98245483+p5quared@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants