Improve observation naming, capture web search and local shell calls#25
Merged
Merged
Conversation
- Name generations "LLM" ("LLM Subagent" in subagent threads) instead of
the raw model name; the model stays on the observation's model attribute.
- Name subagent turn roots "Codex Subagent Turn" so they are
distinguishable from main-thread turns in the trace tree.
- Enrich tool observation names with what was called: the shell command
("exec_command: git status"), the web search query, or the MCP
server/tool split from mcp_tool_call_* events; keep the raw tool name in
codex.tool_name metadata.
- Capture web_search_call response items and web_search_begin/end events
(previously dropped entirely), merging them regardless of rollout order.
- Capture local_shell_call response items for models using the built-in
local shell tool.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019Cj5proPC8sHJmTP5DKmx8
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
LLM(orLLM Subagentinside subagent threads) instead of the raw model name. The model is still recorded on the observation's model attribute, so model breakdowns and cost inference are unaffected.Codex Subagent Turn(previously a second identicalCodex Turn), making subagent activity distinguishable in the trace tree.exec_command: git status(thebash -lcwrapper is stripped, command collapsed to one truncated line)web_search: <query>server.tool— the clean server/tool split taken frommcp_tool_call_begin/endevents instead of the mangledserver__toolfunction namecodex.tool_namemetadata.web_search_callresponse items andweb_search_begin/endevents were silently dropped (the README claimed otherwise). The response item and end event are merged regardless of which is written to the rollout first.local_shell_callresponse items are now captured (the built-in shell tool used by some models) — these had the same gap.Validation
LLMgeneration →web_search: <query>/local_shell: git status/server.toolspans, nestedCodex Subagent Turn→LLM Subagent, with usage, WARNING/ERROR levels, and interruption flags intact.🤖 Generated with Claude Code
https://claude.ai/code/session_019Cj5proPC8sHJmTP5DKmx8
Generated by Claude Code