mmintel/instagram-bot
1
0
Fork 0
Code Issues Pull Requests 4 Actions Packages Projects Releases Wiki Activity

fix(memory): Initialize GoalExecutor singleton with username and validate Qdrant deletes #1

Closed
mmintel wants to merge 1 commits from fix-autonomous-memory-purge into main
pull from: fix-autonomous-memory-purge
merge into: mmintel:main
Conversation 0 Commits 1 Files Changed 2 +10 -5
mmintel commented 2026-04-25 20:14:23 +00:00
Owner
Copy Link

This PR addresses the "memory poisoning paradox" during blank_start.

  1. Bug 1: The Initialization Leak (Stale PathMemory)

    • QNavGraph instantiates GoalExecutor.get_instance(device) during startup. Because bot_username wasn't passed, GoalExecutor instantiated the PathMemory component utilizing the default bot_username="" string.
    • This bound PathMemory to the generic collection goap_paths_v1.
    • blank_start wiped the specific collection goap_paths_v1_marisaundmarc, but since the singleton was using goap_paths_v1, it retained the stale memories.
    • Fix: Pre-initialize GoalExecutor.get_instance(device, username) in bot_flow.py before QNavGraph gets built.

  2. Bug 2: The "Poisoned Memory" Paradox in Qdrant Logs

    • Qdrant's SDK .delete() returns success (200 OK) unconditionally — it doesn't care if the document exists. The delete_point method was blindly logging a successful purge whenever the Unlearn mechanic triggered, regardless of whether the vector had actually been deleted by the agent or by blank_start earlier.
    • Fix: Added a client.retrieve(ids=[point_id]) validation check in qdrant_memory.py's delete_point. If the point was wiped by a global blank_start and doesn't exist anymore, it skips the log.
This PR addresses the "memory poisoning paradox" during `blank_start`. 1. **Bug 1: The Initialization Leak (Stale PathMemory)** - `QNavGraph` instantiates `GoalExecutor.get_instance(device)` during startup. Because `bot_username` wasn't passed, `GoalExecutor` instantiated the `PathMemory` component utilizing the default `bot_username=""` string. - This bound `PathMemory` to the generic collection `goap_paths_v1`. - `blank_start` wiped the specific collection `goap_paths_v1_marisaundmarc`, but since the singleton was using `goap_paths_v1`, it retained the stale memories. - **Fix:** Pre-initialize `GoalExecutor.get_instance(device, username)` in `bot_flow.py` *before* `QNavGraph` gets built. 2. **Bug 2: The "Poisoned Memory" Paradox in Qdrant Logs** - Qdrant's SDK `.delete()` returns success (200 OK) unconditionally — it doesn't care if the document exists. The `delete_point` method was blindly logging a successful purge whenever the Unlearn mechanic triggered, regardless of whether the vector had actually been deleted by the agent or by `blank_start` earlier. - **Fix:** Added a `client.retrieve(ids=[point_id])` validation check in `qdrant_memory.py`'s `delete_point`. If the point was wiped by a global `blank_start` and doesn't exist anymore, it skips the log.
mmintel added 1 commit 2026-04-25 20:14:24 +00:00
fix(memory): Initialize GoalExecutor singleton with username and validate Qdrant deletes b7b8ce9558
mmintel closed this pull request 2026-04-25 20:26:20 +00:00
mmintel deleted branch fix-autonomous-memory-purge 2026-04-25 20:26:20 +00:00

Pull request closed

This pull request cannot be reopened because the branch was deleted.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
mmintel
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: mmintel/instagram-bot#1
Delete Branch "fix-autonomous-memory-purge"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?