Release v3.8.21 - AWS Bedrock ARN parsing fix

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

Description

This PR fixes a critical bug in the AWS Bedrock instrumentation where modelId.split(".") fails when using ARN or cross-region model IDs that contain multiple dots. The issue caused a ValueError: too many values to unpack (expected 2) crash when using model IDs like:

• arn:aws:bedrock:us-east-1:<account_id>:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• us.anthropic.claude-sonnet-4-20250514-v1:0

Changes Made

• Fixed patch_converse_stream (line 110): Replaced (vendor, _) = modelId.split(".") with vendor, _ = parse_vendor_and_model_name_from_model_id(modelId)
• Fixed patch_converse (line 145): Applied the same fix as above
• Version bump: Incremented from 3.8.20 to 3.8.21 for this bug fix release
• Consistency improvement: These functions now use the same robust parsing logic as patch_invoke_model and patch_invoke_model_with_response_stream

Root Cause

The patch_converse_stream and patch_converse functions assumed model IDs could always be split into exactly 2 parts using dots, but ARNs and cross-region model IDs contain multiple dots. The codebase already had a robust parse_vendor_and_model_name_from_model_id function that handles these cases correctly, but these two functions weren't using it.

Testing

⚠️Testing Limitation: Local testing was limited due to missing boto3/botocore dependencies in the development environment. The existing test suite in src/tests/aws_bedrock/test_model_id_parsing.py covers the parser function but couldn't be executed locally.

A test script was created that confirmed:

• ✅ Original modelId.split(".") method fails with "too many values to unpack" error for ARN formats
• ✅ The parse_vendor_and_model_name_from_model_id function handles these cases correctly

Review Checklist

• Critical: Verify fix works with ARN format: arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-3-haiku-20240307-v1:0
• Critical: Verify fix works with cross-region format: us.anthropic.claude-sonnet-4-20250514-v1:0
• Regression testing: Confirm simple model IDs still work: anthropic.claude-3-opus-20240229
• Edge cases: Test the parser with foundation model ARNs and custom model ARNs
• Version: Confirm 3.8.21 is appropriate for this bug fix
• CI: Ensure all tests pass, especially AWS Bedrock related tests

Risk Assessment

🔴 High Priority Review: This touches critical path code for AWS Bedrock instrumentation. While the fix is minimal and follows existing patterns in the codebase, thorough testing is essential to prevent regressions.

---

Link to Devin run: https://app.devin.ai/sessions/9b326f6380504040bf8b711ce512ab94
Requested by: karthik@scale3labs.com

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring