File size: 3,189 Bytes
fedd136
 
 
 
 
df98120
fedd136
 
 
 
ca65aec
9e30ee9
 
b11a5e9
ca65aec
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
---
title: Ghost Malone
emoji: 👻
colorFrom: purple
colorTo: blue
sdk: docker
pinned: false
---

# Ghost Malone: Emotion-Aware MCP System

**📢 Project Announcement:** https://x.com/badtrapazoid/status/1990539436795572375
**🌐 Live Space:** https://huggingface.co/spaces/MCP-1st-Birthday/ghostMalone

A three-server emotional engine built for the MCP ecosystem.

What It Does

Ghost Malone turns raw text into a structured emotional understanding, remembers past patterns, and produces responses that fit the user’s state.
Not a chatbot — a small, disciplined mind.

Architecture (Three Servers)
USER → Orchestrator → Emotion Server → Memory Server → Reflection Server → Output

1. Emotion Server

Russell’s Circumplex (valence + arousal)

Fast pattern matching across 8 affect states

Outputs: labels, valence, arousal, tone

~31ms latency

2. Memory Server

Rolling 50-entry history

Stores text + emotional metadata

Recalls past patterns for personalization

~66ms latency

3. Reflection Server

Claude-driven tone adaptation

Uses emotion + memory + need to shape response

~5.3s latency (dominant cost)

Two Lexicons (Core Intelligence)
Needs Lexicon

5 core needs: autonomy, connection, security, rest, recognition

24 context patterns → 47 inference rules

Aligns emotion with human motive

95.2% accuracy vs BPNSFS scale

Intervention Lexicon

Evidence-based strategies

Constitutional gating:

confidence ≥ 0.70

arousal ≥ 0.40

depth ≥ 2 messages

Prevents overstepping / unsolicited advice

Pipeline (Six Steps)

Emotion analysis

Needs inference

Memory recall + store

Reflection (tone-aware response)

Intervention check

Response assembly

Total latency: ~5.5s.

What Makes It Different
1. Needs, not just emotions

“Sad” branches to different needs (connection vs autonomy vs security).

2. Memory-aware

Responses reference earlier feelings.

3. Constitutional alignment

No forced advice.
No toxic positivity.
User controls sensitivity via sliders.

4. Tunable thresholds

Real-time control of intervention behavior.

5. Emotional trajectory visualization

Simple plot showing how the user is moving on the Circumplex.

Core Example (One Glance)

Input: “I feel so isolated and alone.”

Emotion: sad, lonely (valence -0.6, arousal 0.4)

Need: connection (0.92)

Memory: user mentioned “feeling left out at work”

Response: grounded, gentle reflection

Intervention (if gated): connection strategies


## README Example Section



```markdown
## 🎯 Try These Examples

**For Connection needs:**
1. "I feel so isolated and alone"
2. "Nobody really understands what I'm going through"

**For Autonomy needs:**
1. "I feel so trapped and powerless in this situation"
2. "I have no control over anything anymore"

**For Security needs:**
1. "Everything feels so uncertain and scary"
2. "I'm worried about what's going to happen next"

**For Rest needs:**
1. "I'm so burnt out and exhausted"
2. "I'm completely drained and can't keep going"

**For Recognition needs:**
1. "Nobody notices all the work I do"
2. "I feel completely invisible and unappreciated"

💡 Interventions appear on the 2nd message when thresholds are met.
```