Show which model Codex sessions use

[clkao]

Summary

• Codex sessions now display the model name (e.g. gpt-5-codex, o3-pro) alongside each message, matching what Gemini sessions already show
• Makes it easy to see at a glance which model was used for each turn, including when models change mid-session

How it works

Codex writes a turn_context entry at the start of each user turn containing the active model. The parser now reads this and stamps it on all messages in that turn.

Changes

• Parse turn_context entries from Codex JSONL and extract the model field
• Track current model in the session builder and apply it to all subsequent messages (both text and tool calls)
• Add CodexTurnContextJSON test helper
• Three test cases: model applied to messages, model changes across turns, no turn_context

Test plan

• New unit tests pass (TestParseCodexSession_TurnContextModel)
• All existing Codex parser tests pass (27 tests)
• go vet and go fmt clean
• Verify with real Codex session files containing turn_context entries

🤖 Generated with Claude Code

[roborev-ci]

roborev: Combined Review (a6b177d)

Verdict: The PR successfully propagates model metadata, but requires a
fix to prevent stale model state across turns.

Medium

• Location: internal/parser/codex.go:69
  Problem: currentModel is only updated when turn_context.payload.model is non-empty. A later turn_context that omits or clears model will leave the previous turn's model in place, misattributing all subsequent messages and tool calls.
  Fix: Reset the state on every turn_context entry (e.g., b.currentModel = payload.Get("model").Str)
  and add a regression test for a turn with a missing/empty model following a modeled turn.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)

[roborev-ci]

roborev: Combined Review (24dc878)

Summary Verdict: The code is clean; no issues of medium or higher severity were found.

---

Synthesized from 3 reviews (agents: codex, gemini | types: default, security)