✓ PASS

pydantic-ai 2.61.0

usage.cached-tokens · anthropic

Requirements

Id	Level	Status	Evidence
total-reconciles	MUST	SATISFIED	usage.total=1557 (primary); truth.usage.total=1557
cached-input-reported	SHOULD	SATISFIED	usage.cached_input=1536 (primary); truth.usage.cached_input=1536

What the provider sent

Finish: stop
Raw finish: end_turn
Text: A hot pan helps moisture evaporate so mushrooms brown.
Tool calls: []
Usage: { "alternatives": { "input": { "uncached_input": 0 } }, "cached_input": 1536, "input": 1536, "output": 21, "raw": { "cache_creation_input_tokens": 0, "cache_read_input_tokens": 1536, "input_tokens": 0, "output_tokens": 21 }, "reasoning": null, "total": 1557 }
Stream completed: null
Error: null

Wire excerpt

status: 200
content-type: application/json
body (json):
{
  "id": "msg_01Aq9w938a90dw8q",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "A hot pan helps moisture evaporate so mushrooms brown."
    }
  ],
  "model": "claude-sonnet-5-5",
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "stop_details": null,
  "usage": {
    "input_tokens": 0,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 1536,
    "output_tokens": 21
  }
}

What the library reported

Outcome

returned

Finish

{
  "raw": "end_turn",
  "source": "ModelResponse.finish_reason",
  "surface": "primary",
  "value": "stop"
}

Text

Value: A hot pan helps moisture evaporate so mushrooms brown.
Surface: primary

Tool calls

[]

Usage

{
  "cached_input": 1536,
  "input": 1536,
  "output": 21,
  "reasoning": 0,
  "surface": {
    "cached_input": "primary",
    "input": "primary",
    "output": "primary",
    "reasoning": "primary",
    "total": "primary"
  },
  "total": 1557
}

Stream

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

Error

null

Unsupported reason

null

Full observation JSON

{
  "elapsed_ms": 222.95150067657232,
  "error": null,
  "finish": {
    "raw": "end_turn",
    "source": "ModelResponse.finish_reason",
    "surface": "primary",
    "value": "stop"
  },
  "format": "wiretruth.observation/v1",
  "library": {
    "api": "Agent.run",
    "language": "python",
    "name": "pydantic-ai",
    "version": "2.61.0"
  },
  "notes": [],
  "outcome": "returned",
  "stream": {
    "chunks_seen": null,
    "completed": null,
    "used": false
  },
  "structured": {
    "attempted": false,
    "returned_as_success": false,
    "value": null
  },
  "text": {
    "surface": "primary",
    "value": "A hot pan helps moisture evaporate so mushrooms brown."
  },
  "tool_calls": [],
  "unsupported_reason": null,
  "usage": {
    "cached_input": 1536,
    "input": 1536,
    "output": 21,
    "reasoning": 0,
    "surface": {
      "cached_input": "primary",
      "input": "primary",
      "output": "primary",
      "reasoning": "primary",
      "total": "primary"
    },
    "total": 1557
  }
}

Fixture

documented anthropic/usage.cached-tokens

Provenance	Value
Source	https://platform.claude.com/docs/en/build-with-claude/prompt-caching#tracking-cache-performance
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. Cache field names and accounting are taken verbatim from Tracking cache performance. The original 916-word cooking guide exceeds the documented Sonnet 5.5 minimum of 512 tokens. ASSUMED warm cache covering the repeated request: input_tokens 0, cache_creation_input_tokens 0, cache_read_input_tokens 1536, output_tokens 21. Exact tokenization and this complete response are not documented. No request was sent twice; adaptive thinking is assumed skipped.

Provenance

Value

Source

https://platform.claude.com/docs/en/build-with-claude/prompt-caching#tracking-cache-performance

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. Cache field names and accounting are taken verbatim from Tracking cache performance. The original 916-word cooking guide exceeds the documented Sonnet 5.5 minimum of 512 tokens. ASSUMED warm cache covering the repeated request: input_tokens 0, cache_creation_input_tokens 0, cache_read_input_tokens 1536, output_tokens 21. Exact tokenization and this complete response are not documented. No request was sent twice; adaptive thinking is assumed skipped.

Reproduce

wiretruth serve anthropic/usage.cached-tokens

Plain-library snippet from the shim manifest.

AnthropicModel(model, provider=AnthropicProvider(base_url=WIRETRUTH_BASE_URL_ANTHROPIC, api_key=WIRETRUTH_API_KEY)); host root

Upstream

No matching upstream report.

History

Version	Grade
2.61.0	✓ PASS
2.56.0	✓ PASS