Timmy_Foundation/the-door
15
0
Fork 0
Code Issues 19 Pull Requests 18 Actions Packages Projects Releases Wiki Activity

feat: behavioral pattern detection — frequency, time, withdrawal (closes #133) #146

Closed
Rockachopa wants to merge 2 commits from feat/133-behavioral-detection into main
pull from: feat/133-behavioral-detection
merge into: Timmy_Foundation:main
Conversation 2 Commits 2 Files Changed 2 +472
Rockachopa commented 2026-04-15 16:33:33 +00:00
Owner
Copy Link

Implementation

Detects crisis risk from behavioral patterns, not just message content.

Behavioral Signals

Signal Detection Risk Weight
Frequency spike >2x baseline msgs/hr 0.15
Frequency drop <30% baseline 0.20
Late night 2-5 AM local time 0.10
Session length increasing Span growing over time 0.10
Withdrawal Sustained frequency drop 0.20
Return after absence >48h gap 0.05
Abrupt termination Emotional + short exit 0.20

API

from behavioral_tracker import BehavioralTracker, get_risk_signals

tracker = BehavioralTracker()
tracker.record("session-1", message_length=100, emotional_content=True)
signals = get_risk_signals("session-1")
print(signals.behavioral_score)  # 0-1

Files

  • behavioral_tracker.py — BehavioralTracker class, SessionEvent, BehavioralSignals
  • tests/test_behavioral_tracker.py — 10 tests covering all signal types

Integration

Part of Epic #102 (Multimodal Crisis Detection). Complements content-based crisis_detector.py.

Closes #133

## Implementation Detects crisis risk from behavioral patterns, not just message content. ### Behavioral Signals | Signal | Detection | Risk Weight | |--------|-----------|-------------| | Frequency spike | >2x baseline msgs/hr | 0.15 | | Frequency drop | <30% baseline | 0.20 | | Late night | 2-5 AM local time | 0.10 | | Session length increasing | Span growing over time | 0.10 | | Withdrawal | Sustained frequency drop | 0.20 | | Return after absence | >48h gap | 0.05 | | Abrupt termination | Emotional + short exit | 0.20 | ### API ```python from behavioral_tracker import BehavioralTracker, get_risk_signals tracker = BehavioralTracker() tracker.record("session-1", message_length=100, emotional_content=True) signals = get_risk_signals("session-1") print(signals.behavioral_score) # 0-1 ``` ### Files - `behavioral_tracker.py` — BehavioralTracker class, SessionEvent, BehavioralSignals - `tests/test_behavioral_tracker.py` — 10 tests covering all signal types ### Integration Part of Epic #102 (Multimodal Crisis Detection). Complements content-based `crisis_detector.py`. Closes #133
Rockachopa added 2 commits 2026-04-15 16:33:34 +00:00
feat: behavioral pattern detection — frequency, time, withdrawal (closes #133) 161a9eee5d
test: behavioral pattern detection (#133)
All checks were successful
Sanity Checks / sanity-test (pull_request) Successful in 3s
Details
Smoke Test / smoke (pull_request) Successful in 5s
Details
ba02b63a13
Timmy approved these changes 2026-04-15 23:05:48 +00:00
Timmy left a comment
Owner
Copy Link

Well-structured behavioral pattern detection. The signal types (frequency spike/drop, late-night, withdrawal, abrupt termination) are clinically relevant. Tests are thorough.

Concerns:

  1. CRITICAL — False negative risk in withdrawal detection: The DROP_THRESHOLD of 0.3 (30% of baseline) may miss gradual withdrawal. Someone going from 10 messages/day to 4 messages/day (60% drop) would not trigger at 0.3. Consider a more sensitive threshold like 0.5 for safety, or make it configurable.

  2. Late-night detection uses server time, not user local time: The LATE_NIGHT_START/END (2-5 AM) check uses the timestamp directly but there is no timezone handling. A user in a different timezone would get wrong results. This needs timezone awareness.

  3. In-memory only is correct for privacy but means behavioral baselines reset on restart. Document this limitation.

  4. Global baseline hardcoded: avg_messages_per_hour=5.0 etc. These should be learned from actual user data or be configurable.

  5. Good test coverage including edge cases for abrupt termination and compounding signals.

Approving with the note that the timezone issue (#2) should be addressed before merging to avoid false negatives for users in different timezones.

Well-structured behavioral pattern detection. The signal types (frequency spike/drop, late-night, withdrawal, abrupt termination) are clinically relevant. Tests are thorough. Concerns: 1. **CRITICAL — False negative risk in withdrawal detection**: The DROP_THRESHOLD of 0.3 (30% of baseline) may miss gradual withdrawal. Someone going from 10 messages/day to 4 messages/day (60% drop) would not trigger at 0.3. Consider a more sensitive threshold like 0.5 for safety, or make it configurable. 2. **Late-night detection uses server time, not user local time**: The LATE_NIGHT_START/END (2-5 AM) check uses the timestamp directly but there is no timezone handling. A user in a different timezone would get wrong results. This needs timezone awareness. 3. **In-memory only is correct for privacy** but means behavioral baselines reset on restart. Document this limitation. 4. **Global baseline hardcoded**: avg_messages_per_hour=5.0 etc. These should be learned from actual user data or be configurable. 5. **Good test coverage** including edge cases for abrupt termination and compounding signals. Approving with the note that the timezone issue (#2) should be addressed before merging to avoid false negatives for users in different timezones.
Rockachopa commented 2026-04-16 01:47:35 +00:00
Author
Owner
Copy Link

Closing: Superseded by #149.

Closing: Superseded by #149.
Rockachopa closed this pull request 2026-04-16 01:47:37 +00:00
All checks were successful
Sanity Checks / sanity-test (pull_request) Successful in 3s
Details
Smoke Test / smoke (pull_request) Successful in 5s
Details

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No Reviewers
Timmy
Labels
Clear labels
backend
Backend / API
 batch-pipeline
batch-pipeline label
 content
Content, copy, testimony
 deploy
Deployment / go-live
 epic
Epic / parent issue
 frontend
Frontend / UI
 hardening
Resilience, fallback, security
 infra
Infrastructure, DevOps, deployment
 priority:critical
Must have for launch
 priority:high
Important for launch
 priority:medium
Nice to have at launch
 protocol
Crisis detection protocol
 throughput-10x
throughput-10x label
 token-masterplan
token-masterplan label
 velocity-engine
Auto-generated by velocity engine
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
Rockachopa Timmy allegro antigravity bezalel claude codex-agent ezra gemini google grok hermes kimi manus perplexity sonnet
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Timmy_Foundation/the-door#146
Delete Branch "feat/133-behavioral-detection"

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?