✓ PASS

litellm 1.105.5

truncation.structured-output · anthropic

Requirements

Id	Level	Status	Evidence
no-partial-success	MUST	SATISFIED	structured.attempted=true; structured.returned_as_success=false
cause-is-truncation	MUST	SATISFIED	finish.value="length" (primary: choices[0].finish_reason); truth.finish="length"; error.category=null; structured.returned_as_success=false
text-reported	SHOULD	SATISFIED	text.value="{\"islands\":[{\"name\":\"Κρήτη\",\"sea\":" (primary); truth.text="{\"islands\":[{\"name\":\"Κρήτη\",\"sea\":"
usage-reported	SHOULD	SATISFIED	usage.input=25 (primary); truth.usage.input=25; usage.output=24 (primary); truth.usage.output=24

What the provider sent

Finish: length
Raw finish: max_tokens
Text: {"islands":[{"name":"Κρήτη","sea":
Tool calls: []
Usage: { "alternatives": {}, "cached_input": null, "input": 25, "output": 24, "raw": { "input_tokens": 25, "output_tokens": 24, "output_tokens_details": { "thinking_tokens": 4 } }, "reasoning": 4, "total": 49 }
Stream completed: null
Error: null

Wire excerpt

status: 200
content-type: application/json
body (json):
{
  "id": "msg_01Aq9w938a90dw8q",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "thinking",
      "thinking": "",
      "signature": "EqQBCgIYAhIM1gbcDa9GJwZA2b3hGgxBdjrkzLoky3dl1pkiMOYds..."
    },
    {
      "type": "text",
      "text": "{\"islands\":[{\"name\":\"Κρήτη\",\"sea\":"
    }
  ],
  "model": "claude-sonnet-5-5",
  "stop_reason": "max_tokens",
  "stop_sequence": null,
  "stop_details": null,
  "usage": {
    "input_tokens": 25,
    "output_tokens": 24,
    "output_tokens_details": {
      "thinking_tokens": 4
    }
  }
}

What the library reported

Outcome

returned

Finish

{
  "raw": "length",
  "source": "choices[0].finish_reason",
  "surface": "primary",
  "value": "length"
}

Text

Value: {"islands":[{"name":"Κρήτη","sea":
Surface: primary

Tool calls

[]

Usage

{
  "cached_input": 0,
  "input": 25,
  "output": 24,
  "reasoning": 4,
  "surface": {
    "cached_input": "primary",
    "input": "primary",
    "output": "primary",
    "reasoning": "primary",
    "total": "primary"
  },
  "total": 49
}

Stream

{
  "chunks_seen": null,
  "completed": null,
  "used": false
}

Error

null

Unsupported reason

null

Full observation JSON

{
  "elapsed_ms": 334.956138394773,
  "error": null,
  "finish": {
    "raw": "length",
    "source": "choices[0].finish_reason",
    "surface": "primary",
    "value": "length"
  },
  "format": "wiretruth.observation/v1",
  "library": {
    "api": "litellm.completion",
    "language": "python",
    "name": "litellm",
    "version": "1.105.5"
  },
  "notes": [],
  "outcome": "returned",
  "stream": {
    "chunks_seen": null,
    "completed": null,
    "used": false
  },
  "structured": {
    "attempted": true,
    "returned_as_success": false,
    "value": null
  },
  "text": {
    "surface": "primary",
    "value": "{\"islands\":[{\"name\":\"Κρήτη\",\"sea\":"
  },
  "tool_calls": [],
  "unsupported_reason": null,
  "usage": {
    "cached_input": 0,
    "input": 25,
    "output": 24,
    "reasoning": 4,
    "surface": {
      "cached_input": "primary",
      "input": "primary",
      "output": "primary",
      "reasoning": "primary",
      "total": "primary"
    },
    "total": 49
  }
}

Fixture

documented anthropic/truncation.structured-output

Provenance	Value
Source	https://platform.claude.com/docs/en/build-with-claude/structured-outputs#invalid-outputs
Notes	Documented shapes, not a recording. Message envelope and stop fields follow handling-stop-reasons; content blocks follow thinking/streaming and handle-tool-calls. Model changed to claude-sonnet-5-5; IDs and scenario text are filled in. Token counts are illustrative, filled in to include thinking within the requested output limit, not measured. The native request uses output_config.format. ASSUMED partial Greek JSON prefix and thinking block at max_tokens; the docs describe incomplete structured output but do not print this exact envelope. The output_tokens_details.thinking_tokens breakdown follows thinking-steering.md (https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost#pricing); the thinking count of 4 is filled in, not measured.

Provenance

Value

Source

https://platform.claude.com/docs/en/build-with-claude/structured-outputs#invalid-outputs

Notes

Documented shapes, not a recording. Message envelope and stop fields follow handling-stop-reasons; content blocks follow thinking/streaming and handle-tool-calls. Model changed to claude-sonnet-5-5; IDs and scenario text are filled in. Token counts are illustrative, filled in to include thinking within the requested output limit, not measured. The native request uses output_config.format. ASSUMED partial Greek JSON prefix and thinking block at max_tokens; the docs describe incomplete structured output but do not print this exact envelope. The output_tokens_details.thinking_tokens breakdown follows thinking-steering.md (https://platform.claude.com/docs/en/build-with-claude/thinking-steering-and-cost#pricing); the thinking count of 4 is filled in, not measured.

Reproduce

wiretruth serve anthropic/truncation.structured-output

Plain-library snippet from the shim manifest.

litellm.completion(model='anthropic/' + model, api_base=WIRETRUTH_BASE_URL_ANTHROPIC, api_key=WIRETRUTH_API_KEY); host root, LiteLLM appends /v1/messages

Upstream

No matching upstream report.

History

Version	Grade
1.105.5	✓ PASS
1.105.1	✓ PASS