Runs / pz-61-agent-capture-row-changed · P1 / Inputs & outputs

Stage agent_step · inputs and outputs

step 1 of 1 · pz-61-agent-capture-row-changed · 33ba22f318c0

Recorded with the stage: instrument agent-planner (agent) 1.0.0 · its configuration's fingerprint
sha256:0e5774cac5af…sha256:0e5774cac5af774b0eafa55f39a9997651cd54780b1b8c8d85172eda67d60fac

What is shown: Outputs are stored and shown after checking; inputs from outside the run are fingerprinted, and an input's text is shown only where its step kept it and the text matches its digest.

Differs from pz-59-agent-capture-baseline · 3214a1c5d383

The span context the app reported, as the run recorded it. onetrace does not check it.

  • trace id b728f85bab6e787aaa4373f6e15231c9
  • span id 20fd583e7d0ffc03

Steps inside steps

  • agent_step
    • intake · recorded in a nested run
    • settle · recorded in a nested run
    • nested run pz-62-agent-capture-row-changed-subagent, its head recorded as
      sha256:72e26c8224b3…sha256:72e26c8224b3594d45af79eba68673abcef4583702017eb2dba4522cdb106345
      · open its page

Model-call records

Model-call records, formats model-call-record/1.0 and model-call-record/1.1: each shown as written.

model_call.json

per-step capture: model_call.steps.json

fieldstatevaluereason, as the record gives it
The call
format
what the format says

Written by the writer: this version's constant.

model-call-record/1.1
call.kind
what the format says

self-hosted or hosted.

self-hosted
call.api
what the format says

vLLM: vllm.LLM.generate. transformers: model.forward (prefill) or the writer's decode loop. Hosted: the endpoint path, without query string or host credentials.

vllm.LLM.generate
credential
what the format says

present if a key was passed. Never the key, its digest or any part of it.

absent
model.requested
what the format says

Declared: the model as the caller asked for it. vLLM: the model argument, as repo@revision for a hub snapshot; transformers: the from_pretrained name and revision; hosted: the request's model field.

declared, not observedexample-org/example-model@0123abcexample data, written for the demo; nothing was observed
model.reported
what the format says

Observed: the model the engine or the provider reports.

declared, not observedexample-org/example-modelexample data, written for the demo; nothing was observed
model.files_digest
what the format says

Observed: sha256 of the canonical JSON map from file name to file sha256, computed once per loaded engine.

declared, not observed
sha256:0ee38db7a82c…sha256:0ee38db7a82cfe16e1a1589d687df47932dd9d2068c41cf62c422dc5c929b256
example data, written for the demo; nothing was observed
tokenizer.digest
what the format says

Observed: the manifest digest of the tokenizer files, computed once per loaded tokenizer.

declared, not observed
sha256:8e476026129c…sha256:8e476026129c3fc77aa325a70bef31ed0862a4f90541c5bcecc3cd442290593e
example data, written for the demo; nothing was observed
provider.fingerprint
what the format says

Observed: a build or system fingerprint the provider reports.

not applicableexample data, written for the demo; nothing was observed; self-hosted: there is no provider
engine.name
what the format says

Observed: vllm or transformers.

declared, not observedvllmexample data, written for the demo; nothing was observed
engine.version
what the format says

Observed: the engine's __version__.

declared, not observed0.0.0-exampleexample data, written for the demo; nothing was observed
The runtime
runtime.torch
what the format says

Observed: torch in the calling process.

declared, not observed0.0.0-exampleexample data, written for the demo; nothing was observed
runtime.cuda
what the format says

Observed: torch.cuda in the calling process.

declared, not observed0.0-exampleexample data, written for the demo; nothing was observed
runtime.gpu_name
what the format says

Observed: torch.cuda in the calling process.

declared, not observedExample GPUexample data, written for the demo; nothing was observed
runtime.driver
what the format says

Observed: from the NVIDIA management library.

declared, not observed0.0-exampleexample data, written for the demo; nothing was observed
runtime.compute_capability
what the format says

Observed: torch.cuda in the calling process.

declared, not observed0.0example data, written for the demo; nothing was observed
dtype
what the format says

Declared, then observed. vLLM: model_config.dtype; transformers: the parameters' dtype.

declared, not observedbfloat16example data, written for the demo; nothing was observed
attention_backend
what the format says

Declared, then observed in the calling process. vLLM's worker's own selection is not observed.

not observedexample data, written for the demo; nothing was observed; the worker's own selection is not observed
The batch
batch.size
what the format says

Observed. vLLM: the requests submitted together in one generate call; transformers: input_ids.shape[0].

not observedexample data, written for the demo; nothing was observed; no call was made
batch.executed_size
what the format says

Observed. transformers: input_ids.shape[0], the same as batch.size.

not observedexample data, written for the demo; nothing was observed; no call was made
batch.padded_width
what the format says

Observed. transformers: input_ids.shape[1].

not applicableexample data, written for the demo; nothing was observed; vLLM pads no batch to a width
batch.position
what the format says

Observed: this request's index in the submitted list, or its row index.

not observedexample data, written for the demo; nothing was observed; no call was made
batch.members
what the format says

Observed: for each other request of the batch, in submission order, the digest of its input token ids. Never text, and never token ids.

not observedexample data, written for the demo; nothing was observed; no call was made
batch.fillers
what the format says

Observed: how many members are fillers the writer added.

not observedexample data, written for the demo; nothing was observed; no call was made
batch.batch_invariant
what the format says

Declared, then observed in the calling process: vllm.envs and two configuration effects. The engine worker's own state is never claimed.

declared, not observedonexample data, constructed for the demo; no model was called
The input and output
input.token_ids_digest
what the format says

Observed: the digest of the prompt token ids the writer passed.

not observedexample data, written for the demo; nothing was observed; no call was made
input.request_digest
what the format says

Observed, hosted: the digest of the canonical request body, after the credential and every header are removed.

not applicableexample data, written for the demo; nothing was observed; self-hosted: there is no request body
output.token_ids_digest
what the format says

Observed: the digest of the generated token ids.

not observedexample data, written for the demo; nothing was observed; no call was made
output.text_digest
what the format says

Observed, hosted: the digest of the returned text.

not applicableexample data, written for the demo; nothing was observed; self-hosted: there is no returned text
output.finish
what the format says

Observed: the finish reason; hosted, if the response carries it.

not observedexample data, written for the demo; nothing was observed; no call was made
fingerprints[0] topk_exact over first-token-logprobs, k 20, decimals nonenot observedexample data, written for the demo; nothing was observed; no logprobs were taken
The capture
steps
what the format says

Optional: the digest of the per-step capture output model_call.<n>.steps.json, classed model-generated, when a capture level is taken. Absent otherwise. The record never holds a token id or any text; the per-step token ids live in that separate output.

observed
sha256:b891d1cf41d0…sha256:b891d1cf41d086267ab96ca798215aeb800c8785fe8a6b68872568d8a026c37a
model_call.steps.json
Named by the record, in no group of the schema
engine.settings.engine_coredeclared, not observedin the calling processexample data, constructed for the demo; no model was called
engine.settings.executordeclared, not observeduniexample data, constructed for the demo; no model was called
engine.settings.executor_builtdeclared, not observeduniexample data, constructed for the demo; no model was called
engine.settings.logits_processorsdeclared, not observedonetrace.model_call.capture._l3_vllm.L3Digestsexample data, constructed for the demo; no model was called
engine.settings.model_runnerdeclared, not observedV1pinned to V1 by the adapter: L3 asked for; example data, constructed for the demo; no model was called
sampling.allowed_token_idsdeclared, not observednoneexample data, constructed for the demo; no model was called
sampling.bad_wordsdeclared, not observednoneexample data, constructed for the demo; no model was called
sampling.frequency_penaltydeclared, not observed0.0example data, constructed for the demo; no model was called
sampling.logit_biasdeclared, not observednoneexample data, constructed for the demo; no model was called
sampling.logprobsdeclared, not observed4example data, constructed for the demo; no model was called
sampling.min_tokensdeclared, not observed0example data, constructed for the demo; no model was called
sampling.presence_penaltydeclared, not observed0.0example data, constructed for the demo; no model was called
sampling.repetition_penaltydeclared, not observed1.0example data, constructed for the demo; no model was called
sampling.structured_outputsdeclared, not observednoneexample data, constructed for the demo; no model was called
sampling.temperaturedeclared, not observed0.0example data, constructed for the demo; no model was called
sampling.thinking_token_budgetdeclared, not observednoneexample data, constructed for the demo; no model was called
steps.countobserved6example data, constructed for the demo; no model was called
steps.k_requesteddeclared, not observed4example data, constructed for the demo; no model was called
steps.k_returned_maxobserved4example data, constructed for the demo; no model was called
steps.k_returned_minobserved4example data, constructed for the demo; no model was called
steps.levelobservedL3example data, constructed for the demo; no model was called
steps.methodobservedsha256-tree-512/v0example data, constructed for the demo; no model was called
steps.surplusobserved0example data, constructed for the demo; no model was called

Token timeline

Each decode step's chosen token, as the steps file holds it: token ids, not text. The file is the model's output (classed model-generated); this console decodes it only on a local console, with a tokenizer whose files match the record's tokenizer.digest.

capture level, as the steps file names it: L3

decode stepchosen token id
011
112
213
314
440
541

Also in the steps file, not shown by this console: digests, method, top_k_ids.

Close calls

No close-call flag: no log-probabilities recorded.

model_call.2.json

per-step capture: not taken

fieldstatevaluereason, as the record gives it
The call
format
what the format says

Written by the writer: this version's constant.

model-call-record/1.0
call.kind
what the format says

self-hosted or hosted.

hosted
call.api
what the format says

vLLM: vllm.LLM.generate. transformers: model.forward (prefill) or the writer's decode loop. Hosted: the endpoint path, without query string or host credentials.

chat.completions.create
credential
what the format says

present if a key was passed. Never the key, its digest or any part of it.

present
model.requested
what the format says

Declared: the model as the caller asked for it. vLLM: the model argument, as repo@revision for a hub snapshot; transformers: the from_pretrained name and revision; hosted: the request's model field.

declared, not observedexample-hosted-modelexample data, written for the demo; nothing was observed
model.reported
what the format says

Observed: the model the engine or the provider reports.

not observedexample data, written for the demo; nothing was observed; no call was made
model.files_digest
what the format says

Observed: sha256 of the canonical JSON map from file name to file sha256, computed once per loaded engine.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
tokenizer.digest
what the format says

Observed: the manifest digest of the tokenizer files, computed once per loaded tokenizer.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
provider.fingerprint
what the format says

Observed: a build or system fingerprint the provider reports.

not observedexample data, written for the demo; nothing was observed; no call was made
engine.name
what the format says

Observed: vllm or transformers.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
engine.version
what the format says

Observed: the engine's __version__.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
The runtime
runtime.torch
what the format says

Observed: torch in the calling process.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
runtime.cuda
what the format says

Observed: torch.cuda in the calling process.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
runtime.gpu_name
what the format says

Observed: torch.cuda in the calling process.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
runtime.driver
what the format says

Observed: from the NVIDIA management library.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
runtime.compute_capability
what the format says

Observed: torch.cuda in the calling process.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
dtype
what the format says

Declared, then observed. vLLM: model_config.dtype; transformers: the parameters' dtype.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
attention_backend
what the format says

Declared, then observed in the calling process. vLLM's worker's own selection is not observed.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
The batch
batch.size
what the format says

Observed. vLLM: the requests submitted together in one generate call; transformers: input_ids.shape[0].

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
batch.executed_size
what the format says

Observed. transformers: input_ids.shape[0], the same as batch.size.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
batch.padded_width
what the format says

Observed. transformers: input_ids.shape[1].

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
batch.position
what the format says

Observed: this request's index in the submitted list, or its row index.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
batch.members
what the format says

Observed: for each other request of the batch, in submission order, the digest of its input token ids. Never text, and never token ids.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
batch.fillers
what the format says

Observed: how many members are fillers the writer added.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
batch.batch_invariant
what the format says

Declared, then observed in the calling process: vllm.envs and two configuration effects. The engine worker's own state is never claimed.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
The input and output
input.token_ids_digest
what the format says

Observed: the digest of the prompt token ids the writer passed.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
input.request_digest
what the format says

Observed, hosted: the digest of the canonical request body, after the credential and every header are removed.

not observedexample data, written for the demo; nothing was observed; no call was made
output.token_ids_digest
what the format says

Observed: the digest of the generated token ids.

not observedexample data, written for the demo; nothing was observed; a hosted API does not show it
output.text_digest
what the format says

Observed, hosted: the digest of the returned text.

not observedexample data, written for the demo; nothing was observed; no call was made
output.finish
what the format says

Observed: the finish reason; hosted, if the response carries it.

not observedexample data, written for the demo; nothing was observed; no call was made
fingerprints[0] tolerant over provider-logprobs, k 5, decimals 3not observedexample data, written for the demo; nothing was observed; no logprobs were taken

Field groups and descriptions: model_call_record.schema.json from the onetrace 0.6.0 source package, copied unchanged; the record itself is checked by the installed onetrace.

Inputs

Recorded asWhat's kept
total · externally-sourced

Verified kept text, matches its digest · 3 characters in all

100
sha256:d77ccb69e3f9…sha256:d77ccb69e3f915ed75b4a2db3ff5893fdd0f433e02fdff100441fb843aae585b
· 5 bytes
run:pz-62-agent-capture-row-changed-subagent · operator-authored

not kept

sha256:72e26c8224b3…sha256:72e26c8224b3594d45af79eba68673abcef4583702017eb2dba4522cdb106345
· 0 bytes

A fingerprint shows whether an input changed between runs. This page shows an input's content where its bytes match a stored output of an earlier stage, or where its step kept its text and the text matches its digest.

Outputs

  • model_call.steps.json ·
    sha256:b891d1cf41d0…sha256:b891d1cf41d086267ab96ca798215aeb800c8785fe8a6b68872568d8a026c37a
    · 684 bytes

    Verified

    {
      "digests": [
        "01fc6be9a75534351c7eff4f69c4aea247e3956dd5e9fb410f500084480ad859",
        "380b80e7fd15ba846177fbb22fd2a6885475ec48b78c5fd678ca0261a5612c97",
        "dbea6f8a0a9e202ceb23bbb0c05ec17cded2f84e7600d54875f420e6e311f29f",
        "eb21b01f9fce4a2bcf90d00e3cdeb1da53c2669e1623611e66c5fa3d02696b68",
        "2ea4afc843783e05a2749741bc089b97bd9ac435dbc0fe330ebb230448cdaf59",
        "9dedf004b5b0f831c62321e380137732b3361080e0089ce9930930a3db19251e"
      ],
      "level": "L3",
      "method": "sha256-tree-512/v0",
      "token_ids": [
        "11",
        "12",
        "13",
        "14",
        "40",
        "41"
      ],
      "top_k_ids": [
        [
          "11",
          "9001",
          "9002",
          "9003"
        ],
        [
          "12",
          "9001",
          "9002",
          "9003"
        ],
        [
          "13",
          "9001",
          "9002",
          "9003"
        ],
        [
          "14",
          "9001",
          "9002",
          "9003"
        ],
        [
          "40",
          "9001",
          "9002",
          "9003"
        ],
        [
          "41",
          "9001",
          "9002",
          "9003"
        ]
      ]
    }
  • model_call.json ·
    sha256:9a3712103a2b…sha256:9a3712103a2bffde2d3438f3a69f1cee3d85d33720f131f3c30c5e8e5d2760fc
    · 7691 bytes · the model-call record, as read above

    Verified

    {
      "attention_backend": {
        "source": "example data, written for the demo; nothing was observed; the worker's own selection is not observed",
        "status": "not observed"
      },
      "batch.batch_invariant": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "on"
      },
      "batch.executed_size": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "batch.fillers": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "batch.members": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "batch.padded_width": {
        "source": "example data, written for the demo; nothing was observed; vLLM pads no batch to a width",
        "status": "not applicable"
      },
      "batch.position": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "batch.size": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "call.api": "vllm.LLM.generate",
      "call.kind": "self-hosted",
      "credential": "absent",
      "dtype": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "bfloat16"
      },
      "engine.name": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "vllm"
      },
      "engine.settings.engine_core": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "in the calling process"
      },
      "engine.settings.executor": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "uni"
      },
      "engine.settings.executor_built": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "uni"
      },
      "engine.settings.logits_processors": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "onetrace.model_call.capture._l3_vllm.L3Digests"
      },
      "engine.settings.model_runner": {
        "source": "pinned to V1 by the adapter: L3 asked for; example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "V1"
      },
      "engine.version": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "0.0.0-example"
      },
      "fingerprints": [
        {
          "decimals": "none",
          "form": "topk_exact",
          "k": "20",
          "quantity": "first-token-logprobs",
          "value": {
            "source": "example data, written for the demo; nothing was observed; no logprobs were taken",
            "status": "not observed"
          }
        }
      ],
      "format": "model-call-record/1.1",
      "input.request_digest": {
        "source": "example data, written for the demo; nothing was observed; self-hosted: there is no request body",
        "status": "not applicable"
      },
      "input.token_ids_digest": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "model.files_digest": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "sha256:0ee38db7a82cfe16e1a1589d687df47932dd9d2068c41cf62c422dc5c929b256"
      },
      "model.reported": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "example-org/example-model"
      },
      "model.requested": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "example-org/example-model@0123abc"
      },
      "output.finish": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "output.text_digest": {
        "source": "example data, written for the demo; nothing was observed; self-hosted: there is no returned text",
        "status": "not applicable"
      },
      "output.token_ids_digest": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "provider.fingerprint": {
        "source": "example data, written for the demo; nothing was observed; self-hosted: there is no provider",
        "status": "not applicable"
      },
      "runtime.compute_capability": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "0.0"
      },
      "runtime.cuda": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "0.0-example"
      },
      "runtime.driver": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "0.0-example"
      },
      "runtime.gpu_name": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "Example GPU"
      },
      "runtime.torch": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "0.0.0-example"
      },
      "sampling.allowed_token_ids": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "none"
      },
      "sampling.bad_words": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "none"
      },
      "sampling.frequency_penalty": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "0.0"
      },
      "sampling.logit_bias": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "none"
      },
      "sampling.logprobs": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "4"
      },
      "sampling.min_tokens": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "0"
      },
      "sampling.presence_penalty": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "0.0"
      },
      "sampling.repetition_penalty": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "1.0"
      },
      "sampling.structured_outputs": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "none"
      },
      "sampling.temperature": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "0.0"
      },
      "sampling.thinking_token_budget": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "none"
      },
      "steps": {
        "source": "model_call.steps.json",
        "status": "observed",
        "value": "sha256:b891d1cf41d086267ab96ca798215aeb800c8785fe8a6b68872568d8a026c37a"
      },
      "steps.count": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "observed",
        "value": "6"
      },
      "steps.k_requested": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "declared, not observed",
        "value": "4"
      },
      "steps.k_returned_max": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "observed",
        "value": "4"
      },
      "steps.k_returned_min": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "observed",
        "value": "4"
      },
      "steps.level": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "observed",
        "value": "L3"
      },
      "steps.method": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "observed",
        "value": "sha256-tree-512/v0"
      },
      "steps.surplus": {
        "source": "example data, constructed for the demo; no model was called",
        "status": "observed",
        "value": "0"
      },
      "tokenizer.digest": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "sha256:8e476026129c3fc77aa325a70bef31ed0862a4f90541c5bcecc3cd442290593e"
      }
    }
  • model_call.2.json ·
    sha256:97c5212188b9…sha256:97c5212188b9b0ae0cc275225cef3f3ce635cc9f892006471c049454fb6c79fa
    · 4020 bytes · the model-call record, as read above

    Verified

    {
      "attention_backend": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.batch_invariant": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.executed_size": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.fillers": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.members": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.padded_width": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.position": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "batch.size": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "call.api": "chat.completions.create",
      "call.kind": "hosted",
      "credential": "present",
      "dtype": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "engine.name": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "engine.version": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "fingerprints": [
        {
          "decimals": "3",
          "form": "tolerant",
          "k": "5",
          "quantity": "provider-logprobs",
          "value": {
            "source": "example data, written for the demo; nothing was observed; no logprobs were taken",
            "status": "not observed"
          }
        }
      ],
      "format": "model-call-record/1.0",
      "input.request_digest": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "input.token_ids_digest": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "model.files_digest": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "model.reported": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "model.requested": {
        "source": "example data, written for the demo; nothing was observed",
        "status": "declared, not observed",
        "value": "example-hosted-model"
      },
      "output.finish": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "output.text_digest": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "output.token_ids_digest": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "provider.fingerprint": {
        "source": "example data, written for the demo; nothing was observed; no call was made",
        "status": "not observed"
      },
      "runtime.compute_capability": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "runtime.cuda": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "runtime.driver": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "runtime.gpu_name": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "runtime.torch": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      },
      "tokenizer.digest": {
        "source": "example data, written for the demo; nothing was observed; a hosted API does not show it",
        "status": "not observed"
      }
    }
  • return.json ·
    sha256:3036e8bf5149…sha256:3036e8bf514997c0c20570671c14c5d05fe6a431e1e1ad9b2d931a62df71aabc
    · 17 bytes

    Verified

    {
      "settled": "110"
    }

This step in the other runs

Each other run of P1, newest first, with this step as its own receipt records it; where a run recorded the step more than once, its last attempt. "Same as this run" marks a value equal to this run's, by exact equality. Inputs are not compared across runs: an input that is an earlier step's receipt holds that run's own receipt, so it differs in every run.

RunOutcomeSettingsInstrumentOutputs
pz-63-agent-capture-batch-off · b79bd329b2d2Completed.onetrace.kept_inputs total same as this runagent-planner 1.0.0
sha256:0e5774cac5af…sha256:0e5774cac5af774b0eafa55f39a9997651cd54780b1b8c8d85172eda67d60fac
same as this run
model_call.steps.json
sha256:b891d1cf41d0…sha256:b891d1cf41d086267ab96ca798215aeb800c8785fe8a6b68872568d8a026c37a
same as this run
model_call.json
sha256:ff7afbcb8fa6…sha256:ff7afbcb8fa65b123b2ee602f03144c810430f56bdd0d482025ae7e0e7a9aadc
model_call.2.json
sha256:97c5212188b9…sha256:97c5212188b9b0ae0cc275225cef3f3ce635cc9f892006471c049454fb6c79fa
same as this run
return.json
sha256:3036e8bf5149…sha256:3036e8bf514997c0c20570671c14c5d05fe6a431e1e1ad9b2d931a62df71aabc
same as this run
pz-59-agent-capture-baseline · 3214a1c5d383Completed.onetrace.kept_inputs total same as this runagent-planner 1.0.0
sha256:0e5774cac5af…sha256:0e5774cac5af774b0eafa55f39a9997651cd54780b1b8c8d85172eda67d60fac
same as this run
model_call.steps.json
sha256:6b2e802ce2ef…sha256:6b2e802ce2ef5f0c5bbe648eaf721379fed6b1da575407368d3af6c2a606952e
model_call.json
sha256:f0a2a8a78b96…sha256:f0a2a8a78b9604ed99e1136147feb07c5f092bf0ca49d3909e0d27f138309d8b
model_call.2.json
sha256:97c5212188b9…sha256:97c5212188b9b0ae0cc275225cef3f3ce635cc9f892006471c049454fb6c79fa
same as this run
return.json
sha256:3036e8bf5149…sha256:3036e8bf514997c0c20570671c14c5d05fe6a431e1e1ad9b2d931a62df71aabc
same as this run
Signed by a known recorderYes · demo key (public seed: proves the record wasn't changed since signing, not who signed it) ·
sha256:6e20550ae0dc…sha256:6e20550ae0dc6b4471d54336d09d74d14f3f623b735045eddd053c444ec3b14c
Committed by a known timeNot anchored · no anchor beside the chain

The run, step by step