fix(pydantic): retype wrapper spans as task to stop cost double-counting#312
Merged
Abhijeet Prasad (AbhiPrasad) merged 1 commit intomainfrom Apr 16, 2026
Merged
Conversation
Agent wrapper spans (agent_run, agent_run_sync, agent_run_stream*, agent_to_cli_sync, model_request*) were tagged type=llm and logged the same usage metrics as their nested leaf `chat <model>` span. Experiment aggregations that sum metrics across type=llm spans therefore counted a single provider call twice (wrapper + leaf), inflating reported tokens and cost ~2x for a single-turn agent and more for multi-turn runs. Retag every wrapper span as SpanTypeAttribute.TASK; only the leaf `chat <model>` emitted by _wrap_concrete_model_class stays LLM. _DirectStreamWrapper is used both as wrapper (direct.model_request_stream) and as leaf (Model.request_stream), so it gains a span_type parameter defaulting to LLM; wrapper callers pass TASK explicitly. Test coverage: flip existing wrapper-type assertions to TASK (leaf chat_span assertions stay LLM) and extend the cassette-backed test_agent_run_async with a regression check that exactly one type=llm span exists and that prompt/completion tokens summed across llm spans equal the leaf's values. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Abhijeet Prasad (AbhiPrasad)
approved these changes
Apr 16, 2026
Abhijeet Prasad (AbhiPrasad)
changed the title
colinbennettbrain
added a commit
that referenced
this pull request
Apr 16, 2026
#312 retyped wrapper spans (agent_run, model_request, streaming wrappers) from LLM to TASK, but they kept logging the same prompt_tokens/completion_tokens/tokens as their nested leaf `chat <model>` span. The server derives estimated_cost per-span from tokens+metadata.model (brainstore estimated_cost.rs), rolls up trace totals with `coalesce_add` over every non-scorer span regardless of type (summary.rs accumulate_metrics), and sums experiment-level token/cost over all non-scorer spans without filtering on span_type='llm' (summary.ts experimentScanSpanSummary). Retyping to TASK therefore did not stop double-counting on any of those three axes. Route every wrapper log site through a new `_wrapper_span_metrics` helper that emits only {start, end, duration, optional time_to_first_token}. Leaf `chat <model>` spans (from _wrap_concrete_model_class and _DirectStreamWrapper when span_type=LLM) keep full _extract_response_metrics. `_DirectStreamWrapper` now branches on span_type since it serves as both leaf and wrapper. Delete now-dead `_extract_usage_metrics` and `_extract_stream_usage_metrics`. Flip existing cassette-backed assertions (test_agent_run_async, test_agent_run_sync, test_agent_run_stream_async, test_agent_with_tools, test_agent_run_stream_sync) to assert prompt_tokens / completion_tokens / tokens / prompt_cached_tokens are absent from wrapper spans and present only on the leaf. No cassette re-recording needed -- the change is purely in post-processing. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
colinbennettbrain
added a commit
that referenced
this pull request
Apr 16, 2026
#312 retyped wrapper spans (agent_run, model_request, streaming wrappers) from LLM to TASK, but they kept logging the same prompt_tokens/completion_tokens/tokens as their nested leaf `chat <model>` span. The server derives estimated_cost per-span from tokens+metadata.model (brainstore estimated_cost.rs), rolls up trace totals with `coalesce_add` over every non-scorer span regardless of type (summary.rs accumulate_metrics), and sums experiment-level token/cost over all non-scorer spans without filtering on span_type='llm' (summary.ts experimentScanSpanSummary). Retyping to TASK therefore did not stop double-counting on any of those three axes. Route every wrapper log site through a new `_wrapper_span_metrics` helper that emits only {start, end, duration, optional time_to_first_token}. Leaf `chat <model>` spans (from _wrap_concrete_model_class and _DirectStreamWrapper when span_type=LLM) keep full _extract_response_metrics. `_DirectStreamWrapper` now branches on span_type since it serves as both leaf and wrapper. Delete now-dead `_extract_usage_metrics` and `_extract_stream_usage_metrics`. Flip existing cassette-backed assertions (test_agent_run_async, test_agent_run_sync, test_agent_run_stream_async, test_agent_with_tools, test_agent_run_stream_sync) to assert prompt_tokens / completion_tokens / tokens / prompt_cached_tokens are absent from wrapper spans and present only on the leaf. No cassette re-recording needed -- the change is purely in post-processing.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Agent wrapper spans (agent_run, agent_run_sync, agent_run_stream*, agent_to_cli_sync, model_request*) were tagged type=llm and logged the same usage metrics as their nested leaf
chat <model>span. Experiment aggregations that sum metrics across type=llm spans therefore counted a single provider call twice (wrapper + leaf), inflating reported tokens and cost ~2x for a single-turn agent and more for multi-turn runs.Retag every wrapper span as SpanTypeAttribute.TASK; only the leaf
chat <model>emitted by _wrap_concrete_model_class stays LLM. _DirectStreamWrapper is used both as wrapper (direct.model_request_stream) and as leaf (Model.request_stream), so it gains a span_type parameter defaulting to LLM; wrapper callers pass TASK explicitly.Test coverage: flip existing wrapper-type assertions to TASK (leaf chat_span assertions stay LLM) and extend the cassette-backed test_agent_run_async with a regression check that exactly one type=llm span exists and that prompt/completion tokens summed across llm spans equal the leaf's values.