/continue list previews showed JSON debris (e.g. `{`, `" } ]} === Response ===`)
for some sessions. Two causes in _preview_from_file:
1. Models sometimes emit an unclosed <summary>, so the non-greedy DOTALL regex
pairs it with a far-away </summary> and captures === block headers / JSON
across rounds. Now only the latest <summary> is used, and it is rejected if
it looks like such a cross-segment capture (contains `=== `/`"role"` or
>200 chars) — falling through instead of digging for older summaries.
2. The fallback scanned for the first non-=== line, returning JSON structure
lines. It now uses the last real user prompt via a new _last_user(), which
scans Prompt blocks directly (so response-less/aborted sessions still
preview) and reuses _user_text() to skip tool_result continuations and all
_INJECT_MARKERS (WORKING MEMORY / SYSTEM TIPS / DANGER ...).
Scanned 261 local logs: garbage previews 26 -> 0.
_user_text was misclassifying any prompt whose text block didn't start
with '### [WORKING MEMORY]' as a real user message — but auto-continue
prompts can interleave [SYSTEM TIPS], [System] regenerate triggers,
[DANGER] guard injections, and similar. Treat any prompt carrying a
tool_result block as auto-continue (the natural signal that this is
the next round of an in-flight LLM call), and extend the injection
marker list as a secondary guard for the rare tool_result-less
synthetic prompts.
extract_ui_messages now renders tool_use headers and tool_result fences
into the assistant content using the same string format that agent_loop
yields live, so fold_turns folds restored sessions identically to live
chat. Tool-only response rounds no longer drop the user message, and
Turn 1 also carries a marker so the first turn folds like the rest.
When the user runs '/continue N' in stapp, the agent's in-memory context
is restored, but the UI previously showed only a single '✅ restored' line
— all prior chat bubbles were missing.
This change parses the target session log and reconstructs the
user/assistant message pairs into st.session_state.messages, so reopening
a session feels like the conversation was never interrupted.
* continue_cmd.py: add extract_ui_messages(path)
- parses model_responses log into [{role, content}, ...]
- groups multi-turn LLM calls (prompts whose text starts with the
'### [WORKING MEMORY]' header) into a single assistant bubble,
inserting the existing '**LLM Running (Turn N) ...**' marker so
fold_turns() renders them as collapsible segments.
- two small helpers (_user_text / _assistant_text) keep parsing local.
* stapp.py: in the /continue branch, resolve the target log path BEFORE
calling handle_frontend_command (which snapshots the current log and
would otherwise shift list_sessions indices), then replace
session_state.messages with the reconstructed history on success.
Falls back to the previous behavior for bare /continue or failure.
Co-authored-by: wjl2023 <wjl2023@users.noreply.github.com>
nianyucatfish