langchain - ✅(Solved) Fix Improve BaseMessage.pretty_repr for multimodal content [5 pull requests, 2 comments, 2 participants]

langchain2026-03-30 06:16:34

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

langchain-ai/langchain#36365•Fetched 2026-04-08 01:52:51

View on GitHub

Comments

Participants

Timeline

Reactions

Author

machinelearningprodigy

Participants

gitbalaji

machinelearningprodigy

Timeline (top)

cross-referenced ×4commented ×2labeled ×2referenced ×2

Currently, the pretty_repr method in BaseMessage (used for pretty_print()) only handles string content fully. When a message contains a list of multimodal content blocks (e.g., text and image_url), it defaults to printing the raw list and dictionaries, which is difficult for developers to read during debugging.

I've already developed a fix that adds human-readable formatting for text, image, audio, video, and reasoning blocks.

Error Message

Error Message and Stack Trace (if applicable)

Root Cause

I've already developed a fix that adds human-readable formatting for text, image, audio, video, and reasoning blocks.

Fix Action

Fix / Workaround

This is a bug, not a usage question.
I added a clear and descriptive title that summarizes this issue.
I used the GitHub search to find a similar question and didn't find it.
I am sure that this is a bug in LangChain rather than my code.
The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).
This is not related to the langchain-community package.
I posted a self-contained, minimal, reproducible example. A maintainer can copy it and run it AS IS.

PR fix notes

PR #36364: fix(core): improve message pretty_repr and cleanup output parser

Repository: langchain-ai/langchain
Author: machinelearningprodigy
State: closed | merged: False
Link: https://github.com/langchain-ai/langchain/pull/36364

Description (problem / solution / changelog)

Resolves #36365

Description

This PR addresses two "Developer Experience" (DX) improvements in langchain-core:

Multimodal pretty_repr: Previously, the pretty_repr method used by pretty_print() printed multimodal content (images, audio, video, reasoning) as raw Python dictionaries. I've updated it to iterate through content blocks and format them into a human-readable representation.
Standardization: Renamed the parameter completion to text in OutputParser.parse_with_prompt, resolving a long-standing legacy TODO.

Tagging & Labels

#core #fix #dx #new-contributor

Verification

Manually verified HumanMessage.pretty_print() with a list of multimodal content blocks (images and reasoning). The output is now formatted cleanly instead of showing raw dictionaries.

Changed files

libs/core/langchain_core/messages/base.py (modified, +39/-2)
libs/core/langchain_core/output_parsers/base.py (modified, +3/-4)

PR #36370: fix(core): improve BaseMessage.pretty_repr for multimodal content

Repository: langchain-ai/langchain
Author: Hafiz-Mohammed
State: closed | merged: False
Link: https://github.com/langchain-ai/langchain/pull/36370

Description (problem / solution / changelog)

Description

Fixes #36365

Currently, BaseMessage.pretty_repr only handles string content elegantly. When a message contains a list of multimodal content blocks, it defaults to printing the raw python dict payloads, which is difficult for developers to parse during debugging/logging.

This PR iterates through the dicts and cleanly formats textual representations for text, reasoning, image, image_url, video, and audio block types instead.

Testing

Added three test groups under tests/unit_tests/messages/test_base.py to ensure pretty_repr accurately handles combinations of list[str | dict] layouts without regressions to standard string payloads.

Disclaimer: This contribution was developed with assistance from an AI coding agent.

Changed files

libs/core/langchain_core/messages/base.py (modified, +55/-2)
libs/core/tests/unit_tests/messages/test_base.py (added, +46/-0)

PR #36372: Improve BaseMessage.pretty_repr for multimodal content

Repository: langchain-ai/langchain
Author: hkc5
State: closed | merged: False
Link: https://github.com/langchain-ai/langchain/pull/36372

Description (problem / solution / changelog)

Description

Currently, the pretty_repr method in BaseMessage (used for pretty_print()) only handles string content fully. When a message contains a list of multimodal content blocks (e.g., text and image_url), it defaults to printing the raw list and dictionaries, which is difficult for developers to read during debugging.

Changes

This PR improves the pretty_repr method to handle multimodal content blocks:

Added _format_content_for_pretty_repr helper method to handle list content
Added _format_content_block method to format individual content blocks with human-readable output
Support for the following block types:
- text - Displays the text content directly
- reasoning - Formats with <thinking> tags
- image / image_url - Shows URL, file_id, or base64 indicator with mime_type
- audio - Shows URL, file_id, or base64 indicator with mime_type
- video - Shows URL, file_id, or base64 indicator with mime_type
- file - Shows URL, file_id, or base64 indicator with mime_type
- text-plain - Shows truncated text content
- tool_call - Shows tool name, ID, and arguments
- non_standard - Shows the provider-specific value

Before

After

Backward Compatibility

String content continues to work exactly as before
List content with string elements is handled correctly
Unknown block types fall back to showing the block type and content

Fixes #36365

Changed files

libs/core/langchain_core/messages/base.py (modified, +162/-2)

PR #35402: feat(core): enrich pretty_repr for multimodal content blocks

Repository: langchain-ai/langchain
Author: gitbalaji
State: closed | merged: False
Link: https://github.com/langchain-ai/langchain/pull/35402

Description (problem / solution / changelog)

Summary

Fixes pretty_repr() rendering raw Python list.__repr__() when content is a list (original fix from #34875)
Extends rendering to produce human-readable output for multimodal block types (fixes #36365):
- image_url (OpenAI legacy format): [Image: <url>, detail=<detail>]
- image / video / audio / file: shows URL, file_id, or base64 summary (base64 (N bytes, mime_type))
- reasoning: inline text prefixed with [Reasoning]:
- text-plain: rendered as plain text (same as text)
- Unknown block types still fall back to [type] placeholder

A new private _render_block helper handles single-block dispatch, keeping _render_content_blocks clean.

Why this approach

The existing PR already handled string/text blocks with a _render_content_blocks function. Extending it with a _render_block dispatcher is the minimal change needed — no new public API, no breaking changes, and each block type's rendering is straightforward pattern-matching on the type key.

Areas requiring careful review

The image_url branch handles OpenAI's legacy nested format {"type": "image_url", "image_url": {"url": "..."}} — distinct from the LangChain v1 image block with a top-level url key
The detail field on image_url blocks is forwarded in the rendered string (e.g. [Image: https://..., detail=high])
reasoning blocks show their text inline; empty/missing reasoning field renders as [Reasoning]

Test plan

All 24 unit tests pass (pytest tests/unit_tests/messages/test_base.py)
Existing tests updated to expect [Image: <url>] instead of [image_url]
New tests cover: image_url with/without detail, image with url/file_id/base64/no-source, audio, video, file, reasoning, text-plain
ruff check passes

Note: This PR was developed with AI assistance (Claude Sonnet 4.6 via Claude Code).

🤖 Generated with Claude Code

Changed files

.github/ISSUE_TEMPLATE/bug-report.yml (modified, +2/-0)
.github/ISSUE_TEMPLATE/feature-request.yml (modified, +4/-0)
.github/PULL_REQUEST_TEMPLATE.md (modified, +8/-3)
.github/actions/uv_setup/action.yml (modified, +1/-1)
.github/dependabot.yml (modified, +40/-12)
.github/images/logo-dark.svg (modified, +6/-25)
.github/images/logo-light.svg (modified, +6/-25)
.github/pr-file-labeler.yml (removed, +0/-128)
.github/scripts/check_diff.py (modified, +15/-2)
.github/scripts/get_min_versions.py (modified, +1/-1)
.github/scripts/pr-labeler-config.json (added, +84/-0)
.github/scripts/pr-labeler.js (added, +271/-0)
.github/workflows/_refresh_model_profiles.yml (added, +202/-0)
.github/workflows/_release.yml (modified, +14/-10)
.github/workflows/_test.yml (modified, +2/-2)
.github/workflows/auto-label-by-package.yml (modified, +3/-0)
.github/workflows/check_diffs.yml (modified, +1/-67)
.github/workflows/codspeed.yml (added, +109/-0)
.github/workflows/integration_tests.yml (modified, +2/-2)
.github/workflows/pr_labeler.yml (added, +212/-0)
.github/workflows/pr_labeler_backfill.yml (added, +130/-0)
.github/workflows/pr_labeler_file.yml (removed, +0/-28)
.github/workflows/pr_labeler_title.yml (removed, +0/-44)
.github/workflows/pr_lint.yml (modified, +12/-1)
.github/workflows/refresh_model_profiles.yml (modified, +25/-73)
.github/workflows/require_issue_link.yml (added, +308/-0)
.github/workflows/tag-external-contributions.yml (removed, +0/-148)
.github/workflows/tag-external-issues.yml (added, +205/-0)
.github/workflows/v03_api_doc_build.yml (modified, +5/-2)
.mcp.json (modified, +5/-1)
AGENTS.md (modified, +5/-5)
CLAUDE.md (modified, +5/-5)
CONTRIBUTING.md (removed, +0/-15)
README.md (modified, +43/-35)
libs/README.md (modified, +4/-1)
libs/core/Makefile (modified, +11/-7)
libs/core/langchain_core/_api/deprecation.py (modified, +1/-1)
libs/core/langchain_core/_security/_ssrf_protection.py (modified, +48/-11)
libs/core/langchain_core/cross_encoders.py (added, +18/-0)
libs/core/langchain_core/language_models/base.py (modified, +18/-0)
libs/core/langchain_core/language_models/chat_models.py (modified, +30/-4)
libs/core/langchain_core/language_models/llms.py (modified, +6/-6)
libs/core/langchain_core/messages/base.py (modified, +70/-2)
libs/core/langchain_core/messages/block_translators/openai.py (modified, +50/-0)
libs/core/langchain_core/outputs/chat_result.py (modified, +3/-2)
libs/core/langchain_core/outputs/llm_result.py (modified, +3/-2)
libs/core/langchain_core/runnables/graph_mermaid.py (modified, +5/-0)
libs/core/langchain_core/tracers/base.py (modified, +22/-3)
libs/core/langchain_core/tracers/langchain.py (modified, +30/-7)
libs/core/langchain_core/utils/function_calling.py (modified, +2/-2)
libs/core/langchain_core/utils/pydantic.py (modified, +6/-1)
libs/core/langchain_core/version.py (modified, +1/-1)
libs/core/pyproject.toml (modified, +1/-1)
libs/core/scripts/lint_imports.sh (modified, +3/-2)
libs/core/tests/unit_tests/language_models/chat_models/test_base.py (modified, +147/-0)
libs/core/tests/unit_tests/messages/test_base.py (added, +206/-0)
libs/core/tests/unit_tests/test_ssrf_protection.py (modified, +19/-0)
libs/core/tests/unit_tests/test_tools.py (modified, +17/-0)
libs/core/tests/unit_tests/tracers/test_langchain.py (modified, +149/-21)
libs/core/tests/unit_tests/utils/test_pydantic.py (modified, +19/-0)
libs/core/uv.lock (modified, +116/-118)
libs/langchain/Makefile (modified, +3/-2)
libs/langchain/langchain_classic/agents/mrkl/output_parser.py (modified, +1/-3)
libs/langchain/langchain_classic/agents/output_parsers/react_single_input.py (modified, +1/-3)
libs/langchain/langchain_classic/retrievers/document_compressors/cross_encoder.py (modified, +2/-15)
libs/langchain/pyproject.toml (modified, +4/-4)
libs/langchain/tests/README.md (removed, +0/-3)
libs/langchain/tests/unit_tests/agents/output_parsers/test_react_single_input.py (modified, +32/-0)
libs/langchain/tests/unit_tests/agents/test_mrkl_output_parser.py (modified, +30/-0)
libs/langchain/tests/unit_tests/test_pytest_config.py (modified, +1/-1)
libs/langchain/uv.lock (modified, +343/-292)
libs/langchain_v1/Makefile (modified, +12/-8)
libs/langchain_v1/langchain/__init__.py (modified, +1/-1)
libs/langchain_v1/langchain/agents/factory.py (modified, +43/-7)
libs/langchain_v1/langchain/agents/middleware/__init__.py (modified, +3/-0)
libs/langchain_v1/langchain/agents/middleware/_redaction.py (modified, +19/-1)
libs/langchain_v1/langchain/chat_models/base.py (modified, +2/-0)
libs/langchain_v1/pyproject.toml (modified, +2/-2)
libs/langchain_v1/tests/cassettes/test_inference_to_native_output[True].yaml.gz (modified, +0/-0)
libs/langchain_v1/tests/cassettes/test_inference_to_tool_output[True].yaml.gz (modified, +0/-0)
libs/langchain_v1/tests/cassettes/test_strict_mode[True].yaml.gz (modified, +0/-0)
libs/langchain_v1/tests/unit_tests/agents/middleware/implementations/test_pii.py (modified, +52/-0)
libs/langchain_v1/tests/unit_tests/agents/test_response_format.py (modified, +1/-1)
libs/langchain_v1/tests/unit_tests/chat_models/test_chat_models.py (modified, +1/-1)
libs/langchain_v1/tests/unit_tests/conftest.py (modified, +20/-0)
libs/langchain_v1/uv.lock (modified, +280/-236)
libs/model-profiles/Makefile (modified, +46/-9)
libs/model-profiles/langchain_model_profiles/cli.py (modified, +7/-0)
libs/model-profiles/scripts/lint_imports.sh (added, +18/-0)
libs/model-profiles/tests/unit_tests/test_cli.py (modified, +64/-0)
libs/model-profiles/uv.lock (modified, +108/-108)
libs/partners/anthropic/Makefile (modified, +13/-8)
libs/partners/anthropic/langchain_anthropic/_version.py (modified, +1/-1)
libs/partners/anthropic/langchain_anthropic/chat_models.py (modified, +215/-278)
libs/partners/anthropic/langchain_anthropic/data/_profiles.py (modified, +140/-2)
libs/partners/anthropic/langchain_anthropic/middleware/prompt_caching.py (modified, +121/-13)
libs/partners/anthropic/pyproject.toml (modified, +3/-3)
libs/partners/anthropic/scripts/lint_imports.sh (modified, +3/-2)
libs/partners/anthropic/tests/integration_tests/test_chat_models.py (modified, +29/-12)
libs/partners/anthropic/tests/unit_tests/middleware/test_prompt_caching.py (modified, +233/-200)

PR #36736: fix(core): improve BaseMessage.pretty_repr for multimodal content

Repository: langchain-ai/langchain
Author: slepybear
State: closed | merged: False
Link: https://github.com/langchain-ai/langchain/pull/36736

Description (problem / solution / changelog)

Description

This PR improves BaseMessage.pretty_repr() to handle multimodal content blocks with human-readable output instead of falling back to raw Python dict representations.

Before:

================================ Human Message =================================

[{'type': 'image_url', 'url': 'https://example.com/image.png'}, {'type': 'text', 'text': 'Describe this image'}]

After:

================================ Human Message =================================

[image: https://example.com/image.png]
Describe this image

Changes

Added _format_content() helper function that handles different content block types:
- text blocks: rendered as plain text
- image/image_url blocks: rendered as [image: url] or [image: base64 media_type]
- audio blocks: rendered as [audio: base64 media_type] or [audio]
- video blocks: rendered as [video: base64 media_type] or [video]
- reasoning blocks: rendered as [reasoning: text] or [reasoning]
- Unknown block types: fall back to str(block)
Updated BaseMessage.pretty_repr() to use _format_content() instead of directly printing self.content
Removed the # TODO: handle non-string content. comment since this is now handled
Added 17 unit tests covering all content block types

Testing

Added tests/unit_tests/messages/test_pretty_repr.py with 17 test cases
All existing 215 message tests still pass
ruff check passes with no errors

Fixes #36365

🤖 Generated with Claude Code

Changed files

libs/core/langchain_core/messages/base.py (modified, +66/-2)
libs/core/tests/unit_tests/messages/test_pretty_repr.py (added, +173/-0)

Code Example

from langchain_core.messages import HumanMessage
msg = HumanMessage(content=[
    {"type": "text", "text": "What is in this image?"},
    {"type": "image_url", "image_url": {"url": "https://example.com/image.png"}}
])
# Currently outputs raw Python dictionaries
print("Current Output:")
msg.pretty_print()

---

RAW_BUFFERClick to expand / collapse

Checked other resources

This is a bug, not a usage question.
I added a clear and descriptive title that summarizes this issue.
I used the GitHub search to find a similar question and didn't find it.
I am sure that this is a bug in LangChain rather than my code.
The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).
This is not related to the langchain-community package.
I posted a self-contained, minimal, reproducible example. A maintainer can copy it and run it AS IS.

Package (Required)

Related Issues / PRs

No response

Reproduction Steps / Example Code (Python)

from langchain_core.messages import HumanMessage
msg = HumanMessage(content=[
    {"type": "text", "text": "What is in this image?"},
    {"type": "image_url", "image_url": {"url": "https://example.com/image.png"}}
])
# Currently outputs raw Python dictionaries
print("Current Output:")
msg.pretty_print()

Error Message and Stack Trace (if applicable)

Description

I've already developed a fix that adds human-readable formatting for text, image, audio, video, and reasoning blocks.

Code Example

from langchain_core.messages import HumanMessage

# Current behavior outputs raw dict:
msg = HumanMessage(content=[{"type": "text", "text": "Hi"}, {"type": "image_url", "image_url": {"url": "..."}}])
msg.pretty_print() 
# -> Output: [{'type': 'text', ...}, ...]


### System Info

System Information
------------------
> OS:  Windows
> OS Version:  10.0.26200
> Python Version:  3.13.6 (tags/v3.13.6:4e66535, Aug  6 2025, 14:36:00) [MSC v.1944 64 bit (AMD64)]

Package Information
-------------------

Optional packages not installed
-------------------------------
> langsmith
> deepagents
> deepagents-cli

extent analysis

Fix Plan

To improve the readability of the pretty_print() method for multimodal content blocks, we need to modify the pretty_repr method in BaseMessage to handle different content types.

Here are the steps:

Update the pretty_repr method to check the type of each content block.
Add specific formatting for text, image, audio, video, and reasoning blocks.
Use recursion to handle nested content blocks.

Code Changes

class BaseMessage:
    # ...

    def pretty_repr(self):
        if isinstance(self.content, str):
            return self.content
        elif isinstance(self.content, list):
            formatted_content = []
            for block in self.content:
                if block["type"] == "text":
                    formatted_content.append(f"Text: {block['text']}")
                elif block["type"] == "image_url":
                    formatted_content.append(f"Image: {block['image_url']['url']}")
                # Add more conditions for other content types
                else:
                    formatted_content.append(str(block))
            return "\n".join(formatted_content)
        else:
            return str(self.content)

    def pretty_print(self):
        print(self.pretty_repr())

Verification

To verify the fix, create a HumanMessage with multimodal content and call the pretty_print() method:

msg = HumanMessage(content=[
    {"type": "text", "text": "What is in this image?"},
    {"type": "image_url", "image_url": {"url": "https://example.com/image.png"}}
])
msg.pretty_print()

This should output:

Text: What is in this image?
Image: https://example.com/image.png

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

#api #ssr #model download #tokenizer error #prompt formatting

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

langchain - ✅(Solved) Fix Improve BaseMessage.pretty_repr for multimodal content [5 pull requests, 2 comments, 2 participants]

Recommended Tools

GitHub issue graph ai analysis

Error Message

Error Message and Stack Trace (if applicable)

Root Cause

Fix Action

Fix / Workaround

PR fix notes

PR #36364: fix(core): improve message pretty_repr and cleanup output parser

Description (problem / solution / changelog)

Description

Tagging & Labels

Verification

Changed files

PR #36370: fix(core): improve BaseMessage.pretty_repr for multimodal content

Description (problem / solution / changelog)

Description

Testing

Changed files

PR #36372: Improve BaseMessage.pretty_repr for multimodal content

Description (problem / solution / changelog)

Description

Changes

Before

After

Backward Compatibility

Changed files

PR #35402: feat(core): enrich pretty_repr for multimodal content blocks

Description (problem / solution / changelog)

Summary

Why this approach

Areas requiring careful review

Test plan

Changed files

PR #36736: fix(core): improve BaseMessage.pretty_repr for multimodal content

Description (problem / solution / changelog)

Description

Changes

Testing

Changed files

Code Example

Checked other resources

Package (Required)

Related Issues / PRs

Reproduction Steps / Example Code (Python)

Error Message and Stack Trace (if applicable)

Description

Code Example

extent analysis

Fix Plan

Code Changes

Verification

Still need to ship something?

RELATED_DISCOVERY

TRENDING