fix: use actual model name from token-usage.jsonl in effective tokens footer prefix

[Copilot]

When a user configures a model alias (e.g. agent) in their workflow frontmatter, the ET footer suffix showed age00 — the reduced form of the alias — instead of the real model identifier resolved by the API (e.g. son46 for claude-sonnet-4-6).

Root cause

GH_AW_ENGINE_MODEL stores the user-supplied alias. Footer rendering read that variable for the compact ● <model> <ET> prefix, with no access to the actual model name reported in token-usage.jsonl.

Changes

• parse_mcp_gateway_log.cjs — Added getPrimaryModel(byModel) to pick the model with the highest ET from token-usage.jsonl data; exports it as a new effective_tokens_model step output alongside effective_tokens.

• compiler_main_job.go — Exposes effective_tokens_model as an agent job output.

• compiler_safe_outputs_job.go / notify_comment.go — Passes GH_AW_EFFECTIVE_TOKENS_MODEL env var (from agent job output) to the safe-outputs and failure-notification jobs.

• messages_footer.cjs — Added resolveEffectiveTokensModelName() (GH_AW_EFFECTIVE_TOKENS_MODEL || GH_AW_ENGINE_MODEL); all ET-prefix paths use it. The XML model: marker intentionally keeps GH_AW_ENGINE_MODEL — it records the configured alias.

• effective_tokens.cjs — getEffectiveTokensSuffix() prefers GH_AW_EFFECTIVE_TOKENS_MODEL over GH_AW_ENGINE_MODEL.

• Lock files / golden snapshots — Recompiled 234 workflows; golden files updated with new effective_tokens_model job output.

Result

// Before — user configured model: "agent"
> Generated by [Daily Team Evolution Insights](...) · ● age00 7.1M

// After — actual model from token-usage.jsonl: "claude-sonnet-4-6-20250514"
> Generated by [Daily Team Evolution Insights](...) · ● son46 7.1M

[pelikhan]

@copilot merge main and recompile, fix conflicts

[Copilot]

Pull request overview

This PR fixes effective-token (ET) footer rendering when workflows use a model alias (e.g., agent) by propagating the actual resolved model name from token-usage.jsonl through job outputs/env vars so the compact ● <model> <ET> prefix uses the real model identifier (e.g., son46) instead of the alias-derived identifier.

Changes:

• Add a new effective_tokens_model step/job output sourced from token-usage.jsonl (selecting the model with the highest ET).
• Propagate GH_AW_EFFECTIVE_TOKENS_MODEL into safe-outputs and failure-notification paths and prefer it over GH_AW_ENGINE_MODEL for ET suffix/prefix rendering.
• Recompile/update workflow lockfiles to include the new output and env var wiring.

Show a summary per file

File	Description
actions/setup/js/parse_mcp_gateway_log.cjs	Exports effective_tokens_model based on the primary model from token-usage.jsonl (highest ET).
actions/setup/js/messages_footer.cjs	Prefers GH_AW_EFFECTIVE_TOKENS_MODEL for ET model-prefix rendering (with fallback).
actions/setup/js/effective_tokens.cjs	Uses GH_AW_EFFECTIVE_TOKENS_MODEL (fallback to GH_AW_ENGINE_MODEL) when building the ET suffix.
actions/setup/js/messages.test.cjs	Adds coverage ensuring the ET model prefix prefers the actual model over a configured alias.
pkg/workflow/compiler_main_job.go	Exposes effective_tokens_model as an agent job output.
pkg/workflow/compiler_safe_outputs_job.go	Passes GH_AW_EFFECTIVE_TOKENS_MODEL into the safe-outputs job environment.
pkg/workflow/notify_comment.go	Passes GH_AW_EFFECTIVE_TOKENS_MODEL into failure-notification/conclusion job environment.
.github/workflows/*.lock.yml (many files)	Updates generated workflow lockfiles to include the new effective_tokens_model output and env var propagation.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 251/251 changed files
• Comments generated: 0