litellm - ✅(Solved) Fix [Bug]: KeyError: 'name' in translate_anthropic_tools_to_openai when guardrails process Anthropic built-in tools (bash, text_editor, computer) [1 pull requests, 1 participants]

litellm2026-03-12 17:44:50

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

BerriAI/litellm#23461•Fetched 2026-04-08 00:44:07

View on GitHub

Comments

Participants

Timeline

Reactions

Author

delflein

Participants

delflein

Timeline (top)

referenced ×4labeled ×2subscribed ×2cross-referenced ×1

Error Message

File "/usr/lib/python3.13/site-packages/litellm/proxy/anthropic_endpoints/endpoints.py", line 53, in anthropic_response result = await base_llm_response_processor.base_process_llm_request(...) File "/usr/lib/python3.13/site-packages/litellm/proxy/common_request_processing.py", line 807, in base_process_llm_request self.data, logging_obj = await self.common_processing_pre_call_logic(...) File "/usr/lib/python3.13/site-packages/litellm/proxy/common_request_processing.py", line 667, in common_processing_pre_call_logic self.data = await proxy_logging_obj.pre_call_hook(...) File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1400, in pre_call_hook raise e File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1349, in pre_call_hook result = await self._process_guardrail_callback(...) File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1017, in _process_guardrail_callback response = await self._execute_guardrail_hook(...) File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 892, in _execute_guardrail_hook return await target.async_pre_call_hook(...) File "/usr/lib/python3.13/site-packages/litellm/proxy/guardrails/guardrail_hooks/unified_guardrail/unified_guardrail.py", line 124, in async_pre_call_hook data = await endpoint_translation.process_input_messages(...) File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/chat/guardrail_translation/handler.py", line 79, in process_input_messages LiteLLMAnthropicMessagesAdapter().translate_anthropic_to_openai( anthropic_message_request=cast(AnthropicMessagesRequest, data.copy()) ) File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py", line 910, in translate_anthropic_to_openai new_kwargs["tools"], tool_name_mapping = self.translate_anthropic_tools_to_openai( tools=cast(List[AllAnthropicToolsValues], regular_tools), model=new_kwargs.get("model"), ) File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py", line 736, in translate_anthropic_tools_to_openai original_name = tool["name"] ~~~~^^^^^^^^ KeyError: 'name'

Root Cause

Note: This does NOT happen on /v1/chat/completions (OpenAI format) — only on the Anthropic pass-through endpoint, because only that path receives native Anthropic tool types.

Fix Action

Fixed

Fixed by PR: fix(anthropic): skip built-in tools in translate_anthropic_tools_to_openai (https://github.com/BerriAI/litellm/pull/23473)

PR fix notes

PR #23473: fix(anthropic): skip built-in tools in translate_anthropic_tools_to_openai

Repository: BerriAI/litellm
Author: netbrah
State: closed | merged: False
Link: https://github.com/BerriAI/litellm/pull/23473

Description (problem / solution / changelog)

Summary

Fixes #23461

Anthropic built-in tools (bash_20250124, text_editor_20250124, computer_20250124) only have a type field — no name. When guardrails process requests containing these tools via translate_anthropic_to_openai, the translation crashes with KeyError: 'name' at line 775 of transformation.py.

Root cause: The tool filtering at line 939-943 only separates web search tools. Built-in tools pass _is_web_search_tool and land in regular_tools, which are then passed to translate_anthropic_tools_to_openai where tool["name"] is unconditionally accessed.

Fix: Skip tools without a name key in translate_anthropic_tools_to_openai. These are Anthropic-native tools with no OpenAI equivalent — they should be passed through to the backend untranslated.

Changes

litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py — 2-line guard: if "name" not in tool: continue
Added 2 regression tests: unit-level (translate_anthropic_tools_to_openai with mixed built-in + regular tools) and end-to-end (translate_anthropic_to_openai full request)

Test plan

test_translate_anthropic_tools_to_openai_skips_builtin_tools — verifies built-in tools are skipped, regular tools translate correctly
test_translate_anthropic_to_openai_with_builtin_and_regular_tools — end-to-end with AnthropicMessagesRequest
All existing tests pass (no regressions)

Made with Cursor

Changed files

litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py (modified, +7/-4)
tests/test_litellm/llms/anthropic/experimental_pass_through/adapters/test_anthropic_experimental_pass_through_adapters_transformation.py (modified, +148/-0)

Code Example

// Regular Anthropic tool (has "name") ✅
{"name": "my_tool", "description": "...", "input_schema": {...}}

// Anthropic built-in tools (NO "name") ❌
{"type": "bash_20250124"}
{"type": "text_editor_20250124"}
{"type": "computer_20250124", "display_width_px": 1024, "display_height_px": 768}

---

def translate_anthropic_tools_to_openai(
    self, tools: List[AllAnthropicToolsValues], model: Optional[str] = None
) -> Tuple[List[ChatCompletionToolParam], Dict[str, str]]:
    new_tools: List[ChatCompletionToolParam] = []
    tool_name_mapping: Dict[str, str] = {}
    mapped_tool_params = ["name", "input_schema", "description", "cache_control"]
    for tool in tools:
        original_name = tool["name"]  # 💥 KeyError here for built-in tools
        # ...

---

for tool in tools:
    if "name" not in tool:
        # Built-in Anthropic tool type (bash, text_editor, computer) — skip translation
        continue
    original_name = tool["name"]
    # ...

---

guardrails:
  - guardrail_name: gdpr-eu-business-identifiers
    litellm_params:
      guardrail: litellm_content_filter
      default_on: true

---

curl -X POST https://your-proxy/anthropic/v1/messages \
  -H "x-api-key: sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4.6",
    "max_tokens": 4096,
    "tools": [
      {"type": "bash_20250124"},
      {"type": "text_editor_20250124"},
      {"name": "regular_tool", "description": "A normal tool", "input_schema": {"type": "object"}}
    ],
    "messages": [{"role": "user", "content": "Hello"}]
  }'

---

File "/usr/lib/python3.13/site-packages/litellm/proxy/anthropic_endpoints/endpoints.py", line 53, in anthropic_response
    result = await base_llm_response_processor.base_process_llm_request(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/common_request_processing.py", line 807, in base_process_llm_request
    self.data, logging_obj = await self.common_processing_pre_call_logic(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/common_request_processing.py", line 667, in common_processing_pre_call_logic
    self.data = await proxy_logging_obj.pre_call_hook(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1400, in pre_call_hook
    raise e
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1349, in pre_call_hook
    result = await self._process_guardrail_callback(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1017, in _process_guardrail_callback
    response = await self._execute_guardrail_hook(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 892, in _execute_guardrail_hook
    return await target.async_pre_call_hook(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/guardrails/guardrail_hooks/unified_guardrail/unified_guardrail.py", line 124, in async_pre_call_hook
    data = await endpoint_translation.process_input_messages(...)
File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/chat/guardrail_translation/handler.py", line 79, in process_input_messages
    LiteLLMAnthropicMessagesAdapter().translate_anthropic_to_openai(
        anthropic_message_request=cast(AnthropicMessagesRequest, data.copy())
    )
File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py", line 910, in translate_anthropic_to_openai
    new_kwargs["tools"], tool_name_mapping = self.translate_anthropic_tools_to_openai(
        tools=cast(List[AllAnthropicToolsValues], regular_tools),
        model=new_kwargs.get("model"),
    )
File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py", line 736, in translate_anthropic_tools_to_openai
    original_name = tool["name"]
                    ~~~~^^^^^^^^
KeyError: 'name'

RAW_BUFFERClick to expand / collapse

Disclaimer:

Bug filed with assistance of Claude-Opus-4.6

What happened?

When using Claude Code (or any client that sends Anthropic built-in tool types like bash_20250124, text_editor_20250124, computer_20250124) through the Anthropic pass-through endpoint (/anthropic/v1/messages), the request crashes with a KeyError: 'name' during guardrail pre-call processing.

The crash occurs in translate_anthropic_tools_to_openai() at line 736 of litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py. This function assumes every tool has a "name" field, but Anthropic's built-in tool types only have a "type" field — no "name":

// Regular Anthropic tool (has "name") ✅
{"name": "my_tool", "description": "...", "input_schema": {...}}

// Anthropic built-in tools (NO "name") ❌
{"type": "bash_20250124"}
{"type": "text_editor_20250124"}
{"type": "computer_20250124", "display_width_px": 1024, "display_height_px": 768}

The buggy code (line 736):

def translate_anthropic_tools_to_openai(
    self, tools: List[AllAnthropicToolsValues], model: Optional[str] = None
) -> Tuple[List[ChatCompletionToolParam], Dict[str, str]]:
    new_tools: List[ChatCompletionToolParam] = []
    tool_name_mapping: Dict[str, str] = {}
    mapped_tool_params = ["name", "input_schema", "description", "cache_control"]
    for tool in tools:
        original_name = tool["name"]  # 💥 KeyError here for built-in tools
        # ...

The fix should skip or handle built-in tool types that don't have a "name" field. For example:

for tool in tools:
    if "name" not in tool:
        # Built-in Anthropic tool type (bash, text_editor, computer) — skip translation
        continue
    original_name = tool["name"]
    # ...

Steps to Reproduce

Configure LiteLLM proxy with any guardrail enabled (e.g., litellm_content_filter for GDPR):

guardrails:
  - guardrail_name: gdpr-eu-business-identifiers
    litellm_params:
      guardrail: litellm_content_filter
      default_on: true

Send a request via the Anthropic pass-through endpoint (/anthropic/v1/messages) with built-in tool types:

curl -X POST https://your-proxy/anthropic/v1/messages \
  -H "x-api-key: sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4.6",
    "max_tokens": 4096,
    "tools": [
      {"type": "bash_20250124"},
      {"type": "text_editor_20250124"},
      {"name": "regular_tool", "description": "A normal tool", "input_schema": {"type": "object"}}
    ],
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Observe the KeyError: 'name' crash.

Note: This does NOT happen on /v1/chat/completions (OpenAI format) — only on the Anthropic pass-through endpoint, because only that path receives native Anthropic tool types.

Relevant log output

File "/usr/lib/python3.13/site-packages/litellm/proxy/anthropic_endpoints/endpoints.py", line 53, in anthropic_response
    result = await base_llm_response_processor.base_process_llm_request(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/common_request_processing.py", line 807, in base_process_llm_request
    self.data, logging_obj = await self.common_processing_pre_call_logic(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/common_request_processing.py", line 667, in common_processing_pre_call_logic
    self.data = await proxy_logging_obj.pre_call_hook(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1400, in pre_call_hook
    raise e
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1349, in pre_call_hook
    result = await self._process_guardrail_callback(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 1017, in _process_guardrail_callback
    response = await self._execute_guardrail_hook(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/utils.py", line 892, in _execute_guardrail_hook
    return await target.async_pre_call_hook(...)
File "/usr/lib/python3.13/site-packages/litellm/proxy/guardrails/guardrail_hooks/unified_guardrail/unified_guardrail.py", line 124, in async_pre_call_hook
    data = await endpoint_translation.process_input_messages(...)
File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/chat/guardrail_translation/handler.py", line 79, in process_input_messages
    LiteLLMAnthropicMessagesAdapter().translate_anthropic_to_openai(
        anthropic_message_request=cast(AnthropicMessagesRequest, data.copy())
    )
File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py", line 910, in translate_anthropic_to_openai
    new_kwargs["tools"], tool_name_mapping = self.translate_anthropic_tools_to_openai(
        tools=cast(List[AllAnthropicToolsValues], regular_tools),
        model=new_kwargs.get("model"),
    )
File "/usr/lib/python3.13/site-packages/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py", line 736, in translate_anthropic_tools_to_openai
    original_name = tool["name"]
                    ~~~~^^^^^^^^
KeyError: 'name'

Related issues

#22821 — Presidio guardrail broken with Anthropic native API (same area: guardrail tool translation on pass-through endpoint)
#6427 — Feature request for supporting Anthropic built-in tool types (computer_use, bash, text_editor)

Additional context

Client: Claude Code (uses /anthropic/v1/messages with built-in tools)
LiteLLM version: v1.81.14
Guardrails: Any guardrail that triggers the unified_guardrail → process_input_messages → translate_anthropic_to_openai path
Component: Proxy

extent analysis

Fix Plan

To fix the KeyError: 'name' issue, we need to modify the translate_anthropic_tools_to_openai function to handle Anthropic built-in tool types that don't have a "name" field.

Here are the steps:

Modify the translate_anthropic_tools_to_openai function to check if the "name" field exists in the tool before trying to access it.
If the "name" field does not exist, skip the translation for that tool.

Code Changes

def translate_anthropic_tools_to_openai(
    self, tools: List[AllAnthropicToolsValues], model: Optional[str] = None
) -> Tuple[List[ChatCompletionToolParam], Dict[str, str]]:
    new_tools: List[ChatCompletionToolParam] = []
    tool_name_mapping: Dict[str, str] = {}
    mapped_tool_params = ["name", "input_schema", "description", "cache_control"]
    for tool in tools:
        if "name" not in tool:
            # Built-in Anthropic tool type (bash, text_editor, computer) — skip translation
            continue
        original_name = tool["name"]
        #... rest of the function remains the same

Alternatively, you can also use the .get() method to provide a default value if the "name" field does not exist:

original_name = tool.get("name")
if original_name is None:
    # Built-in Anthropic tool type (bash, text_editor, computer) — skip translation
    continue

Verification

To verify that the fix worked, you can send a request via the Anthropic pass-through endpoint (/anthropic/v1/messages) with built-in tool types and check that the KeyError: 'name' crash no longer occurs.

Extra Tips

Make sure to test the fix with different types of tools, including built-in Anthropic tool types and regular tools with a "name" field.
Consider adding a log message or a warning when a built-in Anthropic tool type is skipped to help with debugging and monitoring.

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 #installation #tensor shape #autograd error #prompt formatting #chain error #conversation history #tool integration

Still need to ship something?

×6

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

Data

Security

Network

Code

UI/UX

Text

System

Multimedia

Protocol

API

Engineering

litellm - ✅(Solved) Fix [Bug]: KeyError: 'name' in translate_anthropic_tools_to_openai when guardrails process Anthropic built-in tools (bash, text_editor, computer) [1 pull requests, 1 participants]

Recommended Tools

GitHub issue graph ai analysis

Error Message

Root Cause

Fix Action

Fixed

PR fix notes

PR #23473: fix(anthropic): skip built-in tools in translate_anthropic_tools_to_openai

Description (problem / solution / changelog)

Summary

Changes

Test plan

Changed files

Code Example

Disclaimer:

What happened?

Steps to Reproduce

Relevant log output

Related issues

Additional context

extent analysis

Fix Plan

Code Changes

Verification

Extra Tips

Still need to ship something?

TRENDING

litellm - ✅(Solved) Fix [Bug]: KeyError: 'name' in translate_anthropic_tools_to_openai when guardrails process Anthropic built-in tools (bash, text_editor, computer) [1 pull requests, 1 participants]

Recommended Tools

GitHub issue graph ai analysis

Error Message

Root Cause

Fix Action

Fixed

PR fix notes

PR #23473: fix(anthropic): skip built-in tools in translate_anthropic_tools_to_openai

Description (problem / solution / changelog)

Summary

Changes

Test plan

Changed files

Code Example

Disclaimer:

What happened?

Steps to Reproduce

Relevant log output

Related issues

Additional context

extent analysis

Fix Plan

Code Changes

Verification

Extra Tips

Still need to ship something?

RELATED_DISCOVERY

TRENDING