google / google/adk-python

AnthropicLlm raises NotImplementedError on empty text parts that ADK itself emits (e.g. code execution with no output)

未关闭
#6,982 1 条评论 0 个 reaction 已指派 1 人 已被 @sanketpatil06 认领 在 GitHub 查看
models
主要语言
Python
星标
21.5k
派生
4k
平均合并
1 天 22 小时
30 天内合并 PR
31

描述

## 🔴 Required Information

**Describe the Bug:**
`AnthropicLlm` converts each `Part` in the conversation into an Anthropic content block. A part that carries only an empty text string (`Part(text='')`) matches none of the branches in `_part_to_message_block` (because `if part.text:` is false for `''`) and falls through to the final `raise NotImplementedError("Not supported yet: ...")`, which fails the whole model call with a misleading error.

ADK produces such parts itself: `code_executors/code_execution_utils.py` writes `Part(text='')` into the content when a code execution returns no output. The LiteLLM path (`lite_llm.py`) simply skips empty text parts, so the same conversation works via `LiteLlm` but fails via the native `AnthropicLlm` adapter.

**Steps to Reproduce:**
1. `pip install google-adk anthropic`
2. Run the snippet under *Minimal Reproduction Code*

**Expected Behavior:**
Empty text parts are skipped (Anthropic rejects empty text blocks anyway), consistent with how the adapter already handles unsupported media parts in assistant turns (warn and skip) and with the LiteLLM path.

**Observed Behavior:**
```text
File ".../google/adk/models/anthropic_llm.py", line 578, in _content_to_message_param
message_block.append(_part_to_message_block(part, sanitizer))
File ".../google/adk/models/anthropic_llm.py", line 557, in _part_to_message_block
raise NotImplementedError(f"Not supported yet: {part}")
NotImplementedError: Not supported yet: media_resolution=None code_execution_result=None executable_code=None file_data=None function_call=None function_response=None inline_data=None text='' thought=None thought_signature=None ...
```

**Environment Details:**

- ADK Library Version (pip show google-adk): 2.8.0
- Desktop OS: macOS (arm64)
- Python Version (python -V): 3.14.7

**Model Information:**

- Are you using LiteLLM: No (native `AnthropicLlm`; the LiteLLM path is not affected)
- Which model is being used: N/A (reproduces in message conversion, no API call needed)

---

## 🟡 Optional Information

**Regression:**
Unknown.

**Logs:**
See *Observed Behavior* above.

**Screenshots / Video:**
N/A

**Additional Context:**
PR incoming shortly (skip parts that carry nothing but empty text in `_content_to_message_param`, plus regression tests). Reported with AI assistance; reproduced and reviewed by me.

**Minimal Reproduction Code:**
```python
from google.genai import types
from google.adk.models.anthropic_llm import content_to_message_param

content = types.Content(
role="user",
parts=[types.Part(text="run it"), types.Part(text="")],
)
print(content_to_message_param(content))
```

**How often has this issue occurred?:**

- Always (100%)

贡献指南

打开贡献指南

评估

这个 Issue 还没有评估数据。

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。