– UNSUPPORTED

any-llm 1.40.2

truncation.structured-output-forced-tool · anthropic

Requirements

Id	Level	Status	Evidence

Notes

any-llm completion maps typed response_format to native output_config.format and has no structured parser for an explicitly forced tool. (https://github.com/mozilla-ai/any-llm/blob/1.40.2/src/any_llm/providers/anthropic/utils.py#L660-L686)

What the provider sent

Finish: length
Raw finish: max_tokens
Text: null
Tool calls: [ { "arguments": { "islands": [ { "name": "Κρήτη" } ] }, "arguments_parsed_ok": true, "arguments_raw": "{\"islands\":[{\"name\":\"Κρήτη\"}]}", "id": "toolu_01A09q90qw90lq917835lq9", "name": "structured_output" } ]
Usage: { "alternatives": {}, "cached_input": null, "input": 25, "output": 24, "raw": { "input_tokens": 25, "output_tokens": 24 }, "reasoning": null, "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": "tool_use",
      "id": "toolu_01A09q90qw90lq917835lq9",
      "name": "structured_output",
      "input": {
        "islands": [
          {
            "name": "Κρήτη"
          }
        ]
      }
    }
  ],
  "model": "claude-haiku-4-5",
  "stop_reason": "max_tokens",
  "stop_sequence": null,
  "stop_details": null,
  "usage": {
    "input_tokens": 25,
    "output_tokens": 24
  }
}

What the library reported

Outcome

unsupported

Finish

{
  "raw": null,
  "source": null,
  "surface": "absent",
  "value": null
}

Text

Value: null
Surface: absent

Tool calls

[]

Usage

{
  "cached_input": null,
  "input": null,
  "output": null,
  "reasoning": null,
  "surface": {
    "cached_input": "absent",
    "input": "absent",
    "output": "absent",
    "reasoning": "absent",
    "total": "absent"
  },
  "total": null
}

Stream

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

Error

null

Unsupported reason

any-llm completion maps typed response_format to native output_config.format and has no structured parser for an explicitly forced tool. https://github.com/mozilla-ai/any-llm/blob/1.40.2/src/any_llm/providers/anthropic/utils.py#L660-L686

Full observation JSON

{
  "elapsed_ms": 1.4513442292809486,
  "error": null,
  "finish": {
    "raw": null,
    "source": null,
    "surface": "absent",
    "value": null
  },
  "format": "wiretruth.observation/v1",
  "library": {
    "api": "any_llm.completion",
    "language": "python",
    "name": "any-llm",
    "version": "1.40.2"
  },
  "notes": [],
  "outcome": "unsupported",
  "stream": {
    "chunks_seen": null,
    "completed": null,
    "used": false
  },
  "structured": {
    "attempted": false,
    "returned_as_success": false,
    "value": null
  },
  "text": {
    "surface": "absent",
    "value": null
  },
  "tool_calls": [],
  "unsupported_reason": "any-llm completion maps typed response_format to native output_config.format and has no structured parser for an explicitly forced tool. https://github.com/mozilla-ai/any-llm/blob/1.40.2/src/any_llm/providers/anthropic/utils.py#L660-L686",
  "usage": {
    "cached_input": null,
    "input": null,
    "output": null,
    "reasoning": null,
    "surface": {
      "cached_input": "absent",
      "input": "absent",
      "output": "absent",
      "reasoning": "absent",
      "total": "absent"
    },
    "total": null
  }
}

Fixture

documented anthropic/truncation.structured-output-forced-tool

Provenance	Value
Source	https://platform.claude.com/docs/en/agents-and-tools/tool-use/define-tools#forcing-tool-use
Notes	Documented shapes, not a recording. Message envelope follows handling-stop-reasons; tool_use object and forced tool choice follow define-tools. Model changed to claude-haiku-4-5; message ID follows the handle-tool-calls example. Scenario data and token counts are filled in, not measured. ASSUMED truncation: input stays an object as documented, but the island is missing its required sea. The docs do not show this exact forced Haiku truncation. No thinking block: forced tool use skips thinking. The call override supplies explicit tools and named tool_choice while output_schema retains the structured intent; no native output_config.format is sent.

Provenance

Value

Source

https://platform.claude.com/docs/en/agents-and-tools/tool-use/define-tools#forcing-tool-use

Notes

Documented shapes, not a recording. Message envelope follows handling-stop-reasons; tool_use object and forced tool choice follow define-tools. Model changed to claude-haiku-4-5; message ID follows the handle-tool-calls example. Scenario data and token counts are filled in, not measured. ASSUMED truncation: input stays an object as documented, but the island is missing its required sea. The docs do not show this exact forced Haiku truncation. No thinking block: forced tool use skips thinking. The call override supplies explicit tools and named tool_choice while output_schema retains the structured intent; no native output_config.format is sent.

Reproduce

wiretruth serve anthropic/truncation.structured-output-forced-tool

Plain-library snippet from the shim manifest.

any_llm.completion(model=model, provider='anthropic', api_base=WIRETRUTH_BASE_URL_ANTHROPIC, api_key=WIRETRUTH_API_KEY); host root, SDK appends /v1/messages and uses x-api-key

Upstream

No matching upstream report.

History

Version	Grade
1.40.2	– UNSUPPORTED
1.38.0	– UNSUPPORTED
1.33.0	– UNSUPPORTED