Single meta_model config + unified agent transcripts - #11
Merged
Conversation
Merge the five per-command meta-agent model keys (draft_model, improve_model, tasks_model, ship_model, check_model) into a single meta_model key. It still defaults to the first models entry and each command's --model flag still overrides it; setting any of the old keys now errors as unknown. Reformat the scaffolded models list to block style, one model per line. Bump to 0.0.2dev. Also render the run's prompt as a leading block in Codex transcripts. Codex's exec --json stream never echoes the prompt, so AgentResult now carries it and render_codex_events prints it, matching what a Claude transcript shows. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the two provider-specific transcript renderers (claude-code-log for Claude, an in-house Codex renderer) with a single path: a run is mapped into a new harness-neutral trajectory model (acumen.trajectory), one mapper per harness (Claude Code, Codex), and one renderer turns any trajectory into HTML. Adding a harness is now one mapper, not a new renderer, and both providers' reports look alike. Each run also writes a portable trajectory.json beside its .html. Claude is rendered in-house too, so claude-code-log is dropped from the claude extra. The renderer collapses tool calls and reasoning into <details> toggles (auto-opened on error), and renders agent text as markdown — bold/italic/inline code, links, headings, lists, fenced code, and GitHub-flavored tables — all escape-first so no transcript content can inject markup. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #11 +/- ##
==========================================
+ Coverage 77.84% 78.44% +0.60%
==========================================
Files 29 30 +1
Lines 4942 5336 +394
==========================================
+ Hits 3847 4186 +339
- Misses 1095 1150 +55
🚀 New features to boost your workflow:
|
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.
Two related changes to how acumen configures its meta-agents and renders their transcripts.
1. Single
meta_modelconfig keyCollapses the five per-command meta-agent model keys —
draft_model,improve_model,tasks_model,ship_model,check_model— into onemeta_modelkey. It still defaults to the firstmodelsentry, and each command's--modelflag still overrides it per run. A config that sets any of the old keys now errors as an unknown key. The scaffoldedmodels:list is also reformatted to block style (one model per line). Bumps the package to0.0.2dev.2. Unified transcript format + one renderer
Replaces the two provider-specific transcript renderers (
claude-code-logfor Claude, an in-house Codex renderer) with a single path:acumen.trajectory).trajectory.jsonbeside its.html.Claude is now rendered in-house too, so
claude-code-logis dropped from theclaudeextra. The renderer:<details>toggles (auto-opened on error),The Codex prompt (which its
exec --jsonstream never echoes) is now carried on the result and shown as a leading block, matching Claude.Testing
trajectory.jsonserialization,<details>toggles, markdown, and GFM tables.trajectory.jsonbeside each.🤖 Generated with Claude Code