feat: deliver proactive sub-session results to opted-in rooms by overcuriousity · Pull Request #232 · overcuriousity/wintermute

overcuriousity · 2026-03-25T08:06:39Z

Summary

Root cause fixed: Proactive sub-session results were silently dropped by broadcast() because thread_id == "default" was explicitly excluded from Matrix routing. Users never saw proactive check-ins despite them running and completing.
Design: No concept of a "default room" exists — routing to the last active session would leak across users. Per-room explicit opt-in is the correct model.
Delivery chain: is_proactive flag threads from sub_session._report() → _QueueItem → broadcast(), which fans out to all opted-in rooms.

Changes

File	Change
`wintermute/infra/thread_config.py`	Add `proactive_enabled` per-thread field (default `False`), `get_proactive_thread_ids()`
`wintermute/interfaces/slash_commands.py`	Add `/proactive [on\|off]` command
`wintermute/core/sub_session.py`	Tag `[PROACTIVE]` results with `is_proactive=True`
`wintermute/core/llm_thread.py`	`is_proactive` on `_QueueItem`, forwarded through queue and broadcast
`wintermute/main.py`	`broadcast()` fans out to all opted-in rooms when `is_proactive=True`
`docs/configuration.md`	Document feature as experimental
`config.yaml.example`	Update proactive config comments

Test plan

Send /proactive on in a Matrix DM → confirm confirmation message received
Wait for or trigger a proactive session → confirm result appears in that room
Rooms without /proactive on receive nothing
Send /proactive off → subsequent proactive sessions no longer delivered
Send /proactive (no args) → shows current on/off status
Web UI debug panel still shows all proactive sessions as before
Non-proactive system events with thread_id="default" still do not reach Matrix

Note: Proactive sessions run in full mode and can take autonomous actions (spawn workers, create tasks, etc.) — not just summarise. Users opted in will see the LLM-processed result of whatever the session did.

🤖 Generated with Claude Code

Proactive sessions were completing successfully but their results were silently dropped at the Matrix routing layer (broadcast() skipped any thread_id == "default"). Users never saw proactive check-ins. The fix threads an is_proactive flag from sub_session._report() through the LLM queue item to the broadcast closure, which then delivers to all rooms that have opted in via the new /proactive slash command. Changes: - thread_config: add proactive_enabled per-thread field (default: False) and get_proactive_thread_ids() helper - slash_commands: add /proactive [on|off] command to opt rooms in/out - sub_session: tag results from [PROACTIVE] objectives with is_proactive - llm_thread: add is_proactive to _QueueItem, forward through enqueue_system_event() and broadcast call - main: broadcast() delivers to all opted-in rooms when is_proactive=True - docs + config.yaml.example: document feature as experimental Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

This PR fixes delivery of proactive sub-session results by introducing an explicit “opt-in per room/thread” routing path, so proactive outputs (which otherwise originate from thread_id="default") can be fanned out to opted-in Matrix/Signal rooms without leaking to arbitrary “last active” rooms.

Changes:

Add per-thread proactive_enabled config and a helper to list opted-in thread IDs.
Add /proactive [on|off] slash command to toggle opt-in for a room and report status.
Thread an is_proactive flag from sub-session reporting → LLM queue → broadcast(), and fan out proactive results to opted-in rooms.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
`wintermute/main.py`	Adds `is_proactive` to `broadcast()` and proactive fan-out routing to opted-in rooms
`wintermute/interfaces/slash_commands.py`	Adds `/proactive` command for opt-in/out and lists it in `/commands`
`wintermute/infra/thread_config.py`	Introduces `proactive_enabled` and `get_proactive_thread_ids()`
`wintermute/core/sub_session.py`	Tags sub-session reports as proactive based on objective prefix
`wintermute/core/llm_thread.py`	Adds `is_proactive` to queue items and forwards it through system-event broadcasting
`docs/configuration.md`	Documents opt-in delivery and deprecates `proactive_target_thread_id`
`config.yaml.example`	Updates comments to reflect new opt-in behavior

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

- slash_commands: use exact token match instead of startswith to avoid false matches like /proactiveXYZ - sub_session: add explicit is_proactive flag to TaskNode/SubSessionState instead of parsing objective string prefix (brittle coupling removed) - scheduler: pass is_proactive=True when spawning proactive sessions - main: fan-out sends now run concurrently via asyncio.gather; web SSE broadcast moved outside the loop so it fires once regardless of how many rooms are opted in - docs: clarify that proactive_target_thread_id != "default" bypasses room opt-in fan-out entirely (previous wording was misleading) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

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

Copilot · 2026-03-25T10:51:33Z

        if wi and thread_id:
            await wi.broadcast(text, thread_id, reasoning=reasoning)
+        # Proactive results: deliver to all rooms that have opted in via /proactive on.
+        # This fires only when thread_id == "default" (proactive sessions have no specific
+        # target room), ensuring regular messages are never duplicated.
+        if is_proactive and thread_id == "default":
+            opted_in = thread_config_mgr.get_proactive_thread_ids()
+            if opted_in:
+                async def _send_to_room(tid: str) -> None:
+                    try:
+                        if si and tid.startswith("sig_"):
+                            await si.send_message(text, tid)
+                        elif mx and not tid.startswith("web_") and not tid.startswith("sig_"):
+                            await mx.send_message(text, tid)
+                    except Exception:  # noqa: BLE001
+                        logger.warning("Failed to deliver proactive message to %s", tid)
+
+                await asyncio.gather(*(_send_to_room(tid) for tid in opted_in))
+                if wi:
+                    await wi.broadcast(text, thread_id, reasoning=reasoning)


broadcast() currently calls wi.broadcast() twice for proactive messages (is_proactive=True with thread_id == "default"): once unconditionally for any thread_id and again inside the proactive fan-out block. This will duplicate proactive results in the web UI. Consider broadcasting to the web interface only once (e.g., skip the first wi.broadcast when is_proactive and thread_id == "default", or remove the second call and instead add web-thread fan-out if needed).

Copilot · 2026-03-25T10:51:34Z

+                    except Exception:  # noqa: BLE001
+                        logger.warning("Failed to deliver proactive message to %s", tid)
+
+                await asyncio.gather(*(_send_to_room(tid) for tid in opted_in))
+                if wi:
+                    await wi.broadcast(text, thread_id, reasoning=reasoning)


Rooms opting in from the web UI will set proactive_enabled on web_... thread IDs, and get_proactive_thread_ids() will return them, but the proactive fan-out explicitly skips web_ IDs and never calls wi.broadcast() for those target thread IDs. Either filter out web_ thread IDs from proactive opt-in (e.g., reject /proactive on for web threads or exclude them in get_proactive_thread_ids()), or add a wi.broadcast(text, tid, ...) branch so web threads can actually receive proactive messages.

Suggested change

except Exception: # noqa: BLE001

logger.warning("Failed to deliver proactive message to %s", tid)

await asyncio.gather(*(_send_to_room(tid) for tid in opted_in))

if wi:

await wi.broadcast(text, thread_id, reasoning=reasoning)

elif wi and tid.startswith("web_"):

await wi.broadcast(text, tid, reasoning=reasoning)

except Exception: # noqa: BLE001

logger.warning("Failed to deliver proactive message to %s", tid)

await asyncio.gather(*(_send_to_room(tid) for tid in opted_in))

Copilot · 2026-03-25T10:51:34Z

+    # proactive_target_thread_id: "default"  # Deprecated: use /proactive on in each room instead
+    # Proactive delivery (experimental): use the /proactive on slash command in any Matrix/Signal
+    # room to opt that room in to receiving proactive check-in results. Multiple rooms can be
+    # opted in independently. Use /proactive off to remove a room. No config change required.


proactive_target_thread_id is still read by the scheduler and, per the docs in this PR, can be used to force delivery to a specific thread ID (bypassing room opt-in). Labeling it as "Deprecated" here is misleading; consider rewording to indicate it’s optional/advanced and that /proactive on is the recommended default behavior when it remains set to "default".

Suggested change

# proactive_target_thread_id: "default" # Deprecated: use /proactive on in each room instead

# Proactive delivery (experimental): use the /proactive on slash command in any Matrix/Signal

# room to opt that room in to receiving proactive check-in results. Multiple rooms can be

# opted in independently. Use /proactive off to remove a room. No config change required.

# proactive_target_thread_id: "default" # Advanced/optional: keep "default" to respect room-level /proactive on opt-in (recommended), or set a specific thread ID to force delivery there.

# Proactive delivery (experimental): use the /proactive on slash command in any Matrix/Signal

# room to opt that room in to receiving proactive check-in results. Multiple rooms can be

# opted in independently. Use /proactive off to remove a room. No config change required for

# the recommended default behavior (proactive_target_thread_id = "default").

Copilot AI review requested due to automatic review settings March 25, 2026 08:06

Copilot started reviewing on behalf of overcuriousity March 25, 2026 08:07 View session

Copilot AI reviewed Mar 25, 2026

View reviewed changes

Comment thread wintermute/main.py Outdated

Comment thread docs/configuration.md Outdated

Comment thread wintermute/interfaces/slash_commands.py Outdated

Comment thread wintermute/core/sub_session.py Outdated

Comment thread wintermute/main.py Outdated

overcuriousity requested a review from Copilot March 25, 2026 10:46

Copilot started reviewing on behalf of overcuriousity March 25, 2026 10:47 View session

Copilot AI reviewed Mar 25, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: deliver proactive sub-session results to opted-in rooms#232

feat: deliver proactive sub-session results to opted-in rooms#232
overcuriousity wants to merge 2 commits into
mainfrom
feat/proactive-room-delivery

overcuriousity commented Mar 25, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Copilot AI Mar 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

overcuriousity commented Mar 25, 2026

Summary

Changes

Test plan

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 25, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants