PR #36298: fix(core): forward detail field in convert_to_openai_image_block

• Repository: langchain-ai/langchain
• Author: insilications
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36298

Description (problem / solution / changelog)

convert_to_openai_image_block constructs the image_url dict with only the url key, ignoring any detail field in the input block.

This also breaks convert_to_openai_data_block for the Responses API, which checks chat_completions_block["image_url"].get("detail") — a condition that can never be true since the upstream function never includes detail.

Extract detail from the block (top-level, extras, or metadata) and include it in the output when present. Fix the tests covering both functions and both APIs.

Fixes #36297

Social handles

Twitter: @insilications

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +18/-18)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +323/-80)
• libs/core/tests/unit_tests/test_messages.py (modified, +62/-1)

PR #36327: core: Fix convert_to_openai_image_block dropping detail field

• Repository: langchain-ai/langchain
• Author: joaquinhuigomez
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36327

Description (problem / solution / changelog)

Fixes #36297

convert_to_openai_image_block constructs the image_url dict with only the url key, silently dropping the detail field that controls image resolution. This affects both Chat Completions and Responses API paths.

The fix extracts detail from the input block (checking direct attributes, extras, and metadata for backwards compatibility) and includes it in the output dict when present. Added tests covering all lookup paths for both convert_to_openai_image_block and convert_to_openai_data_block, plus no-detail cases to prevent regression.

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +25/-6)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +89/-0)

PR #36337: fix(core): preserve detail parameter in openai image blocks

• Repository: langchain-ai/langchain
• Author: Hafiz-Mohammed
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36337

Description (problem / solution / changelog)

Fixes #36297

This PR updates convert_to_openai_image_block to correctly preserve the optional detail parameter and forward it into the final image_url dictionary. I also added explicit test cases mimicking the reporter's examples to the OpenAI block translator test suite to prevent future regressions.

Added 6 new parameterized test cases to libs/core/tests/unit_tests/messages/block_translators/test_openai.py and successfully ran make test, make format, and make lint against the langchain_core package.

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +24/-6)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +73/-1)

PR #36345: fix(core): forward detail field in convert_to_openai_image_block

• Repository: langchain-ai/langchain
• Author: MDSAMIULSAMI
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36345

Description (problem / solution / changelog)

The convert_to_openai_image_block function silently dropped the optional detail field from image content blocks, affecting both the OpenAI Chat Completions and Responses APIs.

This commit adds a private _extract_detail helper that checks for detail at the top level, in extras, and in metadata (for backwards compatibility). The extracted value is included in the image_url dict when present.

Fixes #36297

Fixes #

Read the full contributing guidelines: https://docs.langchain.com/oss/python/contributing/overview

All contributions must be in English. See the language policy.

If you paste a large clearly AI generated description here your PR may be IGNORED or CLOSED!

Thank you for contributing to LangChain! Follow these steps to have your pull request considered as ready for review.

1. PR title: Should follow the format: TYPE(SCOPE): DESCRIPTION

• Examples:
  • fix(anthropic): resolve flag parsing error
  • feat(core): add multi-tenant support
  • test(openai): update API usage tests
• Allowed TYPE and SCOPE values: https://github.com/langchain-ai/langchain/blob/master/.github/workflows/pr_lint.yml#L15-L33

1. PR description:

• Write 1-2 sentences summarizing the change.
• The Fixes #xx line at the top is required for external contributions — update the issue number and keep the keyword. This links your PR to the approved issue and auto-closes it on merge.
• If there are any breaking changes, please clearly describe them.
• If this PR depends on another PR being merged first, please include "Depends on #PR_NUMBER" in the description.

1. Run make format, make lint and make test from the root of the package(s) you've modified.

• We will not consider a PR unless these three are passing in CI.

1. How did you verify your code works?

Additional guidelines:

• All external PRs must link to an issue or discussion where a solution has been approved by a maintainer, and you must be assigned to that issue. PRs without prior approval will be closed.
• PRs should not touch more than one package unless absolutely necessary.
• Do not update the uv.lock files or add dependencies to pyproject.toml files (even optional ones) unless you have explicit permission to do so by a maintainer.

Social handles (optional)

LinkedIn: https://www.linkedin.com/in/md-samiul-islam-17738a1b9

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +36/-6)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +76/-0)
• libs/core/tests/unit_tests/test_messages.py (modified, +78/-0)

PR #36381: fix(core): preserve detail field in convert_to_openai_image_block

• Repository: langchain-ai/langchain
• Author: IgnazioDS
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36381

Description (problem / solution / changelog)

Summary

Fixes #36297 — convert_to_openai_image_block silently drops the detail field ("low", "high", "auto") from image content blocks.

Root cause: Both the url and base64 branches built image_url with only {"url": ...}, never forwarding detail. The downstream convert_to_openai_data_block Responses API path then conditionally copies detail from the already-built Chat Completions block — so the field was lost for both API targets.

Fix: Extract detail from the input block (top-level key → extras → metadata fallback chain) and conditionally include it in the image_url dict.

# Before
return {"type": "image_url", "image_url": {"url": block["url"]}}

# After
detail = block.get("detail") or block.get("extras", {}).get("detail") or block.get("metadata", {}).get("detail")
image_url: dict[str, Any] = {"url": block["url"]}
if detail:
    image_url["detail"] = detail
return {"type": "image_url", "image_url": image_url}

The same fix applies to the base64 branch.

Test plan

• test_convert_to_openai_image_block_preserves_detail — url path, base64 path, extras-nested detail, absent detail
• test_convert_to_openai_data_block_image_preserves_detail — Chat Completions API and Responses API
• Run pytest libs/core/tests/unit_tests/messages/block_translators/test_openai.py

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +9/-12)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +53/-0)

PR #36425: fix(core): forward detail field in convert_to_openai_image_block

• Repository: langchain-ai/langchain
• Author: Saad-Azi
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36425

Description (problem / solution / changelog)

Why

convert_to_openai_image_block constructs the image_url dict with only the url key, silently dropping any detail field present in the input block. Since convert_to_openai_data_block delegates to convert_to_openai_image_block and checks chat_completions_block['image_url'].get('detail') to forward detail to the Responses API format, the detail field was lost for both the Chat Completions API and the Responses API.

The detail parameter controls image analysis resolution ('low', 'high', 'auto') and is documented by OpenAI. Silently dropping it causes models to silently fall back to their default behavior, which may result in lower-quality or more expensive image analysis than the caller intended.

Fixes #36297

What changed

libs/core/langchain_core/messages/block_translators/openai.py

In convert_to_openai_image_block: resolve detail from the input block (top-level → extras → metadata for backwards compat, following the existing pattern used throughout the file for ilename) and include it in the constructed image_url dict when present. No change needed in convert_to_openai_data_block — it already forwards detail from the inner dict.

No public API changes. Fully backward compatible.

Tests

19 new parametrized unit tests in est_openai.py:

All 19 pass. All pre-existing tests in the file continue to pass.

Areas requiring review

• The detail resolution chain uses or short-circuit: lock.get('detail') or (block.get('extras') or {}).get('detail') or (block.get('metadata') or {}).get('detail'). This matches the existing pattern for ilename lookup in convert_to_openai_data_block for file blocks. If detail could ever be a falsy value (it can't per OpenAI spec — valid values are 'low', 'high', 'auto'), an explicit is not None check would be safer, but this mirrors existing style.

AI disclosure

This PR was developed with AI assistance. I have reviewed every line of the fix and the tests, understand the root cause, and verified all tests pass locally.

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +15/-6)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +179/-0)

PR #36468: fix(core): forward detail field in convert_to_openai_image_block

• Repository: langchain-ai/langchain
• Author: hanxujoy88
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36468

Description (problem / solution / changelog)

Summary

Fixes #36297

The detail field controls image analysis resolution (low, high, auto) in OpenAI API. Previously, this field was silently dropped when converting image blocks via convert_to_openai_image_block.

Changes

• Extract detail from top-level, extras, or metadata fields
• Forward the detail field to the output image_url dict

Testing

Tested manually with all combinations:

• URL with detail at top level ✅
• URL with detail in extras ✅
• URL with detail in metadata ✅
• Base64 with detail ✅
• No detail (backward compat) ✅
• convert_to_openai_data_block for chat/completions API ✅
• convert_to_openai_data_block for responses API ✅

Example

Before:

block = {'type': 'image', 'url': 'https://...', 'detail': 'high'}
convert_to_openai_image_block(block)
# {'type': 'image_url', 'image_url': {'url': 'https://...'}}  # detail lost!

After:
block = {'type': 'image', 'url': 'https://...', 'detail': 'high'}
convert_to_openai_image_block(block)
# {'type': 'image_url', 'image_url': {'url': 'https://...', 'detail': 'high'}}  # detail preserved!

## Changed files

- `libs/core/langchain_core/messages/block_translators/openai.py` (modified, +15/-6)


---

# PR #36469: fix(core): forward detail field in convert_to_openai_image_block

- Repository: langchain-ai/langchain
- Author: hanxujoy88
- State: closed | merged: False
- Link: https://github.com/langchain-ai/langchain/pull/36469

## Description (problem / solution / changelog)

## Summary

Fixes #36297

The `detail` field controls image analysis resolution (`low`, `high`, `auto`) in OpenAI API. Previously, this field was silently dropped when converting image blocks via `convert_to_openai_image_block`.

## Changes

- Extract `detail` from top-level, `extras`, or `metadata` fields
- Forward the `detail` field to the output `image_url` dict

## Testing

Tested manually with all combinations:
- URL with detail at top level ✅
- URL with detail in extras ✅  
- URL with detail in metadata ✅
- Base64 with detail ✅
- No detail (backward compat) ✅
- `convert_to_openai_data_block` for chat/completions API ✅
- `convert_to_openai_data_block` for responses API ✅

## Example

Before:
```python
block = {'type': 'image', 'url': 'https://...', 'detail': 'high'}
convert_to_openai_image_block(block)
# {'type': 'image_url', 'image_url': {'url': 'https://...'}}  # detail lost!

After:

block = {'type': 'image', 'url': 'https://...', 'detail': 'high'}
convert_to_openai_image_block(block)
# {'type': 'image_url', 'image_url': {'url': 'https://...', 'detail': 'high'}}  # detail preserved!

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +15/-6)

PR #36505: fix(core): forward detail field in convert_to_openai_image_block

• Repository: langchain-ai/langchain
• Author: akhilc08
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36505

Description (problem / solution / changelog)

Summary

• convert_to_openai_image_block was silently dropping the detail field when constructing the image_url dict for OpenAI. The detail parameter controls image analysis resolution ("low", "high", "auto") and is important for controlling token usage and analysis quality.
• Added a _get_detail() helper that extracts detail from the block's top-level, extras, or metadata dicts.
• The fix applies to both the Chat Completions API path ({"type": "image_url", "image_url": {"url": "...", "detail": "high"}}) and the Responses API path ({"type": "input_image", "image_url": "...", "detail": "high"}). The Responses API path in convert_to_openai_data_block already had code to forward detail from the intermediate chat completions block, but it was always empty before this fix.
• Added comprehensive tests covering detail from all three source locations, both API paths, and the absence case.

Closes #36297

Test plan

• Added test_convert_to_openai_data_block_detail_field covering:
  • Chat Completions: detail at top level (url and base64), in extras, in metadata, and absent
  • Responses API: detail at top level (url and base64), and absent
• AST validation passes on both modified files

🤖 Generated with Claude Code

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +24/-6)
• libs/core/tests/unit_tests/messages/block_translators/test_openai.py (modified, +94/-0)

PR #36549: Fix convert_to_openai_image_block dropping detail field

• Repository: langchain-ai/langchain
• Author: Bahtya
• State: closed | merged: False
• Link: https://github.com/langchain-ai/langchain/pull/36549

Description (problem / solution / changelog)

Summary

Fixes #36297

convert_to_openai_image_block was dropping the detail field from image content blocks when converting to OpenAI format, for both URL and base64 source types.

Fix

Forward the detail field to the output dict when present in the input block.

Before

return {
    "type": "image_url",
    "image_url": {"url": block["url"]},
}

After

result = {
    "type": "image_url",
    "image_url": {"url": block["url"]},
}
if "detail" in block:
    result["image_url"]["detail"] = block["detail"]
return result

Same pattern applied to the base64 path.

Changed files

• libs/core/langchain_core/messages/block_translators/openai.py (modified, +8/-2)