intentmind 0.2.0__tar.gz → 0.2.1__tar.gz

{intentmind-0.2.0/src/intentmind.egg-info → intentmind-0.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: intentmind
-Version: 0.2.0
+Version: 0.2.1
 Summary: Language-agnostic associative memory runtime for persistent AI systems
 Author: Intentmind Contributors
 License: Intentmind Fair-Code License (Non-Commercial)
@@ -76,7 +76,7 @@ If you are building an autonomous system or an AI agent that lives for more than
 
 ## 🌟 The Solution: Why Intentmind?
 
-Intentmind doesn't just store documents; it builds a **living, breathing Cognitive Graph**. 
+Intentmind doesn't just store documents; it builds a **dynamic associative memory graph with energy-based activation, decay, reinforcement, and traceable recall paths**. 
 
 ### 1. Associative Recall (Semantic Drift Recovery)
 When data is ingested, Intentmind extracts core concepts (nodes) and connects them (edges). When a user mentions a concept, the system doesn't just do a vector search—it **traverses the graph**. 
@@ -107,15 +107,17 @@ GraphRAG is fantastic for static document analysis, but it fails for **persisten
 
 ## 📊 Performance Benchmarks
 
-Intentmind is built to survive long-term deployments where other systems degrade.
+*Synthetic Semantic Drift Benchmark*
+- **Dataset:** 1,000 memory facts
+- **Noise:** 30%
+- **Query type:** delayed indirect recall
 
-| Benchmark Metric | Vanilla RAG | GraphRAG | Intentmind |
-| :--- | :--- | :--- | :--- |
-| **Delayed Recall Accuracy** | Low | Medium | **High** |
-| **Semantic Drift Resistance** | Poor | Medium | **Excellent** |
-| **Memory Pollution (Noise) Resistance** | Fails at scale | Medium | **High** (Auto-pruning) |
-| **Long-session Degradation** | Severe | Moderate | **Near-Zero** |
-| **Contradiction Resolution** | Low | Low | **High** (Energy-based priority) |
+| System | Accuracy |
+| :--- | :--- |
+| Vanilla vector search | 42% |
+| **Intentmind** | **68%** |
+
+*Note: These are local regression benchmarks, not independent research claims. Intentmind is built to be resilient to long-session degradation where other systems degrade.*
 
 ---
 
@@ -141,14 +143,20 @@ pip install "intentmind[all]"
 
 ### 1-Minute Integration
 
-Intentmind is completely language-agnostic.
+While the core graph logic is language-agnostic, extraction quality is **embedding-model dependent**. We recommend using multilingual models for non-English data.
 
 ```python
 from intentmind import IntentmindMemory
 from intentmind.embeddings import SentenceTransformerEmbedder
 
-# 1. Initialize the Cognitive Engine
-mem = IntentmindMemory(embedder=SentenceTransformerEmbedder())
+# Option A: Local deterministic demo (fastest, no LLM needed)
+mem = IntentmindMemory(is_test=True, core_extractor="deterministic")
+
+# Option B: Production embeddings with deterministic extraction
+# mem = IntentmindMemory(embedder=SentenceTransformerEmbedder(), core_extractor="deterministic")
+
+# Option C: Production with full LLM-assisted extraction (requires OPENAI_API_KEY)
+# mem = IntentmindMemory(embedder=SentenceTransformerEmbedder(), core_extractor="llm")
 
 # 2. Teach it concepts (Happens naturally during conversation)
 mem.add("I need car insurance but I have no money.")
@@ -186,7 +194,7 @@ mem = IntentmindMemory.load("my_agent_brain.json")
 
 ## 📊 Visualizing the "Brain"
 
-Intentmind isn't a black box. You can export the entire Cognitive Graph to an interactive HTML map to literally see what your AI is thinking and how it connects concepts.
+Intentmind isn't a black box. You can export the entire Cognitive Graph to an interactive HTML map to visualize the graph structure, energy levels, and synaptic connections.
 
 ```python
 mem.visualize("memory_map.html")

{intentmind-0.2.0 → intentmind-0.2.1}/README.md

@@ -20,7 +20,7 @@ If you are building an autonomous system or an AI agent that lives for more than
 
 ## 🌟 The Solution: Why Intentmind?
 
-Intentmind doesn't just store documents; it builds a **living, breathing Cognitive Graph**. 
+Intentmind doesn't just store documents; it builds a **dynamic associative memory graph with energy-based activation, decay, reinforcement, and traceable recall paths**. 
 
 ### 1. Associative Recall (Semantic Drift Recovery)
 When data is ingested, Intentmind extracts core concepts (nodes) and connects them (edges). When a user mentions a concept, the system doesn't just do a vector search—it **traverses the graph**. 
@@ -51,15 +51,17 @@ GraphRAG is fantastic for static document analysis, but it fails for **persisten
 
 ## 📊 Performance Benchmarks
 
-Intentmind is built to survive long-term deployments where other systems degrade.
+*Synthetic Semantic Drift Benchmark*
+- **Dataset:** 1,000 memory facts
+- **Noise:** 30%
+- **Query type:** delayed indirect recall
 
-| Benchmark Metric | Vanilla RAG | GraphRAG | Intentmind |
-| :--- | :--- | :--- | :--- |
-| **Delayed Recall Accuracy** | Low | Medium | **High** |
-| **Semantic Drift Resistance** | Poor | Medium | **Excellent** |
-| **Memory Pollution (Noise) Resistance** | Fails at scale | Medium | **High** (Auto-pruning) |
-| **Long-session Degradation** | Severe | Moderate | **Near-Zero** |
-| **Contradiction Resolution** | Low | Low | **High** (Energy-based priority) |
+| System | Accuracy |
+| :--- | :--- |
+| Vanilla vector search | 42% |
+| **Intentmind** | **68%** |
+
+*Note: These are local regression benchmarks, not independent research claims. Intentmind is built to be resilient to long-session degradation where other systems degrade.*
 
 ---
 
@@ -85,14 +87,20 @@ pip install "intentmind[all]"
 
 ### 1-Minute Integration
 
-Intentmind is completely language-agnostic.
+While the core graph logic is language-agnostic, extraction quality is **embedding-model dependent**. We recommend using multilingual models for non-English data.
 
 ```python
 from intentmind import IntentmindMemory
 from intentmind.embeddings import SentenceTransformerEmbedder
 
-# 1. Initialize the Cognitive Engine
-mem = IntentmindMemory(embedder=SentenceTransformerEmbedder())
+# Option A: Local deterministic demo (fastest, no LLM needed)
+mem = IntentmindMemory(is_test=True, core_extractor="deterministic")
+
+# Option B: Production embeddings with deterministic extraction
+# mem = IntentmindMemory(embedder=SentenceTransformerEmbedder(), core_extractor="deterministic")
+
+# Option C: Production with full LLM-assisted extraction (requires OPENAI_API_KEY)
+# mem = IntentmindMemory(embedder=SentenceTransformerEmbedder(), core_extractor="llm")
 
 # 2. Teach it concepts (Happens naturally during conversation)
 mem.add("I need car insurance but I have no money.")
@@ -130,7 +138,7 @@ mem = IntentmindMemory.load("my_agent_brain.json")
 
 ## 📊 Visualizing the "Brain"
 
-Intentmind isn't a black box. You can export the entire Cognitive Graph to an interactive HTML map to literally see what your AI is thinking and how it connects concepts.
+Intentmind isn't a black box. You can export the entire Cognitive Graph to an interactive HTML map to visualize the graph structure, energy levels, and synaptic connections.
 
 ```python
 mem.visualize("memory_map.html")

{intentmind-0.2.0 → intentmind-0.2.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "intentmind"
-version = "0.2.0"
+version = "0.2.1"
 description = "Language-agnostic associative memory runtime for persistent AI systems"
 readme = "README.md"
 requires-python = ">=3.10"

{intentmind-0.2.0 → intentmind-0.2.1}/src/intentmind/builders/prompt_builder.py

@@ -50,7 +50,7 @@ class PromptBuilder:
     #  PUBLIC API                                                         #
     # ================================================================== #
 
-    def build(self, user_query, recall_result, emotional_state):
+    def build(self, user_query, recall_result, cognitive_state):
         role = self._build_role()
         graph = self._build_graph_state(recall_result)
         memories = self._build_memories(recall_result)

intentmind-0.2.1/src/intentmind/engines/__init__.py

@@ -0,0 +1,5 @@
+from .intent_engine import IntentEngine
+from .cognitive_state_engine import CognitiveStateEngine
+from .recall_engine import RecallEngine
+from .energy_engine import EnergyEngine
+__all__ = ["IntentEngine", "CognitiveStateEngine", "RecallEngine", "EnergyEngine"]

intentmind-0.2.1/src/intentmind/engines/cognitive_state_engine.py

@@ -0,0 +1,36 @@
+from __future__ import annotations
+
+from ..models import CognitiveState
+from ..embeddings import BaseEmbedder, cosine_similarity
+
+
+class CognitiveStateEngine:
+    def __init__(self, embedder: BaseEmbedder):
+        self.embedder = embedder
+        self.prototypes = {
+            "neutral":    embedder.embed("I am providing or requesting information."),
+            "curious":   embedder.embed("I am curious and want to explore this idea."),
+            "urgent": embedder.embed("This is critical, urgent, and needs immediate attention."),
+            "confident":   embedder.embed("This makes sense and is solid."),
+            "skeptical":   embedder.embed("I am unsure and this needs verification."),
+            "cautious":   embedder.embed("This is risky and requires careful attention."),
+        }
+        self.override_priority = ["urgent", "cautious", "skeptical", "curious", "confident", "neutral"]
+        self.override_threshold = 0.80
+
+    def detect(self, message_embedding) -> CognitiveState:
+        scores = {emotion: cosine_similarity(message_embedding, proto) for emotion, proto in self.prototypes.items()}
+        ranked = sorted(scores.items(), key=lambda x: x[1], reverse=True)
+        top_emotion, top_score = ranked[0]
+        second_score = ranked[1][1]
+        delta = top_score - second_score
+        for priority_emotion in self.override_priority[:2]:
+            if scores[priority_emotion] > self.override_threshold:
+                top_emotion = priority_emotion
+                top_score = scores[priority_emotion]
+                break
+        if top_score < 0.55 or delta < 0.07:
+            state = "neutral"
+        else:
+            state = top_emotion
+        return CognitiveState(current=state, confidence=round(top_score, 3), weak_echo=state in ["curious", "urgent"])

{intentmind-0.2.0 → intentmind-0.2.1}/src/intentmind/engines/recall_engine.py

@@ -11,7 +11,7 @@ class RecallEngine:
         self.store = store
         self.intent_match_threshold = intent_match_threshold
 
-    def activation_score(self, query_embedding, intent, emotional_state):
+    def activation_score(self, query_embedding, intent, cognitive_state):
         query_similarity = cosine_similarity(query_embedding, intent.embedding)
         edges = self.store.get_neighbors(intent.intent_id)
         edge_energy = sum(e.energy for _, e in edges) / len(edges) if edges else 0.40
@@ -28,11 +28,11 @@ class RecallEngine:
             - intent.noise_score * 0.20
         )
         multipliers = {
-            "nötr": 1.00, "merak": 1.00, "heyecan": 1.00,
-            "güven": 1.15, "şüphe": 1.10, "korku": 1.20,
+            "neutral": 1.00, "curious": 1.00,
+            "confident": 1.15, "skeptical": 1.10, "urgent": 1.25, "cautious": 1.20,
         }
-        m = multipliers.get(emotional_state.current, 1.00)
-        factor = 1.0 + ((m - 1.0) * emotional_state.confidence)
+        m = multipliers.get(cognitive_state.current, 1.00)
+        factor = 1.0 + ((m - 1.0) * cognitive_state.confidence)
         return round(base * factor, 4)
 
     def chunk_score(self, query_embedding, chunk, intent, path_strength):
@@ -58,7 +58,7 @@ class RecallEngine:
     def build_layers(
         self,
         query_embedding,
-        emotional_state,
+        cognitive_state,
         query_intent_labels=None,
         query_token_embeddings=None,
         field_seed_ids=None,
@@ -78,7 +78,7 @@ class RecallEngine:
         for intent in self.store.intents.values():
             if intent.state == "archived":
                 continue
-            score = self.activation_score(query_embedding, intent, emotional_state)
+            score = self.activation_score(query_embedding, intent, cognitive_state)
 
             # Query intent matching is resolved once up front. This avoids
             # scanning every query-token embedding against every graph node.
@@ -175,7 +175,7 @@ class RecallEngine:
                     })
         layers[1] = self._dedupe_layer(layers[1])[:12]
 
-        if emotional_state.weak_echo:
+        if cognitive_state.weak_echo:
             seen_l1 = {item["intent"].intent_id for item in layers[1]}
             seen_all = seen_l0 | seen_l1
             
@@ -248,7 +248,7 @@ class RecallEngine:
         self,
         query,
         query_embedding,
-        emotional_state,
+        cognitive_state,
         query_intent_labels=None,
         query_token_embeddings=None,
         energy_engine=None,
@@ -259,7 +259,7 @@ class RecallEngine:
         query_match_intent_ids = self._matching_intent_ids(query_intent_labels, query_token_embeddings)
         layers = self.build_layers(
             query_embedding,
-            emotional_state,
+            cognitive_state,
             query_intent_labels,
             query_token_embeddings,
             query_match_intent_ids=query_match_intent_ids,
@@ -379,7 +379,7 @@ class RecallEngine:
         candidates = sorted(self.remove_duplicates(candidates), key=lambda x: x["score"], reverse=True)
         direct = self._select_direct(candidates)
         associated = self._select_associated(candidates, direct, query_intent_set)
-        weak_echo = [c for c in candidates if c["layer"] == 3][:2] if emotional_state.weak_echo else []
+        weak_echo = [c for c in candidates if c["layer"] == 3][:2] if cognitive_state.weak_echo else []
         
         return {
             "direct_memories": direct,

{intentmind-0.2.0 → intentmind-0.2.1}/src/intentmind/models.py

@@ -111,8 +111,8 @@ class IntentEdge:
 
 
 @dataclass
-class EmotionalState:
-    current: str = "nötr"
+class CognitiveState:
+    current: str = "neutral"
     confidence: float = 1.0
     weak_echo: bool = False
 

{intentmind-0.2.0 → intentmind-0.2.1}/src/intentmind/runtime.py

@@ -3,7 +3,7 @@ from __future__ import annotations
 from pathlib import Path
 from .store import MemoryStore
 from .embeddings import BaseEmbedder, FakeEmbedder
-from .engines import IntentEngine, EmotionEngine, RecallEngine, EnergyEngine
+from .engines import IntentEngine, CognitiveStateEngine, RecallEngine, EnergyEngine
 from .builders import PromptBuilder
 from .persistence import JsonPersistence
 
@@ -61,7 +61,7 @@ class IntentmindMemory:
             model=model,
             extractor=extractor,
         )
-        self._emotion = EmotionEngine(self.embedder)
+        self._cognitive_state = CognitiveStateEngine(self.embedder)
         self._recall = RecallEngine(self._store)
         self._energy = EnergyEngine(self._store)
         self._prompt = PromptBuilder()
@@ -77,8 +77,8 @@ class IntentmindMemory:
         query_emb = self.embedder.embed(user_query)
         t_embed = _time.perf_counter()
 
-        emotional_state = self._emotion.detect(query_emb)
-        t_emotion = _time.perf_counter()
+        cognitive_state = self._cognitive_state.detect(query_emb)
+        t_state = _time.perf_counter()
 
         query_token_embeddings = self._intent_engine.extract_query_intents(user_query)
         t_extract = _time.perf_counter()
@@ -88,14 +88,14 @@ class IntentmindMemory:
         recall_result = self._recall.recall(
             query=user_query, 
             query_embedding=query_emb, 
-            emotional_state=emotional_state,
+            cognitive_state=cognitive_state,
             query_intent_labels=query_intent_labels,
             query_token_embeddings=query_token_embeddings,
             energy_engine=self._energy,
         )
         t_recall = _time.perf_counter()
 
-        prompt = self._prompt.build(user_query=user_query, recall_result=recall_result, emotional_state=emotional_state)
+        prompt = self._prompt.build(user_query=user_query, recall_result=recall_result, cognitive_state=cognitive_state)
         t_prompt = _time.perf_counter()
 
         self.metadata["total_queries"] = self.metadata.get("total_queries", 0) + 1
@@ -103,8 +103,8 @@ class IntentmindMemory:
         total_ms = (t_prompt - t0) * 1000
         latency_breakdown = {
             "embed_query_ms": round((t_embed - t0) * 1000, 2),
-            "emotion_ms": round((t_emotion - t_embed) * 1000, 2),
-            "extractor_ms": round((t_extract - t_emotion) * 1000, 2),
+            "cognitive_state_ms": round((t_state - t_embed) * 1000, 2),
+            "extractor_ms": round((t_extract - t_state) * 1000, 2),
             "recall_ms": round((t_recall - t_extract) * 1000, 2),
             "prompt_ms": round((t_prompt - t_recall) * 1000, 2),
             "total_ms": round(total_ms, 2),
@@ -112,7 +112,7 @@ class IntentmindMemory:
 
         return {
             "prompt": prompt,
-            "emotion": emotional_state.to_dict(),
+            "cognitive_state": cognitive_state.to_dict(),
             "memories": {
                 "direct": len(recall_result["direct_memories"]),
                 "associated": len(recall_result["associated_memories"]),

{intentmind-0.2.0 → intentmind-0.2.1/src/intentmind.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: intentmind
-Version: 0.2.0
+Version: 0.2.1
 Summary: Language-agnostic associative memory runtime for persistent AI systems
 Author: Intentmind Contributors
 License: Intentmind Fair-Code License (Non-Commercial)
@@ -76,7 +76,7 @@ If you are building an autonomous system or an AI agent that lives for more than
 
 ## 🌟 The Solution: Why Intentmind?
 
-Intentmind doesn't just store documents; it builds a **living, breathing Cognitive Graph**. 
+Intentmind doesn't just store documents; it builds a **dynamic associative memory graph with energy-based activation, decay, reinforcement, and traceable recall paths**. 
 
 ### 1. Associative Recall (Semantic Drift Recovery)
 When data is ingested, Intentmind extracts core concepts (nodes) and connects them (edges). When a user mentions a concept, the system doesn't just do a vector search—it **traverses the graph**. 
@@ -107,15 +107,17 @@ GraphRAG is fantastic for static document analysis, but it fails for **persisten
 
 ## 📊 Performance Benchmarks
 
-Intentmind is built to survive long-term deployments where other systems degrade.
+*Synthetic Semantic Drift Benchmark*
+- **Dataset:** 1,000 memory facts
+- **Noise:** 30%
+- **Query type:** delayed indirect recall
 
-| Benchmark Metric | Vanilla RAG | GraphRAG | Intentmind |
-| :--- | :--- | :--- | :--- |
-| **Delayed Recall Accuracy** | Low | Medium | **High** |
-| **Semantic Drift Resistance** | Poor | Medium | **Excellent** |
-| **Memory Pollution (Noise) Resistance** | Fails at scale | Medium | **High** (Auto-pruning) |
-| **Long-session Degradation** | Severe | Moderate | **Near-Zero** |
-| **Contradiction Resolution** | Low | Low | **High** (Energy-based priority) |
+| System | Accuracy |
+| :--- | :--- |
+| Vanilla vector search | 42% |
+| **Intentmind** | **68%** |
+
+*Note: These are local regression benchmarks, not independent research claims. Intentmind is built to be resilient to long-session degradation where other systems degrade.*
 
 ---
 
@@ -141,14 +143,20 @@ pip install "intentmind[all]"
 
 ### 1-Minute Integration
 
-Intentmind is completely language-agnostic.
+While the core graph logic is language-agnostic, extraction quality is **embedding-model dependent**. We recommend using multilingual models for non-English data.
 
 ```python
 from intentmind import IntentmindMemory
 from intentmind.embeddings import SentenceTransformerEmbedder
 
-# 1. Initialize the Cognitive Engine
-mem = IntentmindMemory(embedder=SentenceTransformerEmbedder())
+# Option A: Local deterministic demo (fastest, no LLM needed)
+mem = IntentmindMemory(is_test=True, core_extractor="deterministic")
+
+# Option B: Production embeddings with deterministic extraction
+# mem = IntentmindMemory(embedder=SentenceTransformerEmbedder(), core_extractor="deterministic")
+
+# Option C: Production with full LLM-assisted extraction (requires OPENAI_API_KEY)
+# mem = IntentmindMemory(embedder=SentenceTransformerEmbedder(), core_extractor="llm")
 
 # 2. Teach it concepts (Happens naturally during conversation)
 mem.add("I need car insurance but I have no money.")
@@ -186,7 +194,7 @@ mem = IntentmindMemory.load("my_agent_brain.json")
 
 ## 📊 Visualizing the "Brain"
 
-Intentmind isn't a black box. You can export the entire Cognitive Graph to an interactive HTML map to literally see what your AI is thinking and how it connects concepts.
+Intentmind isn't a black box. You can export the entire Cognitive Graph to an interactive HTML map to visualize the graph structure, energy levels, and synaptic connections.
 
 ```python
 mem.visualize("memory_map.html")

{intentmind-0.2.0 → intentmind-0.2.1}/src/intentmind.egg-info/SOURCES.txt

@@ -21,7 +21,7 @@ src/intentmind/embeddings/base.py
 src/intentmind/embeddings/fake.py
 src/intentmind/embeddings/sentence_transformer.py
 src/intentmind/engines/__init__.py
-src/intentmind/engines/emotion_engine.py
+src/intentmind/engines/cognitive_state_engine.py
 src/intentmind/engines/energy_engine.py
 src/intentmind/engines/intent_engine.py
 src/intentmind/engines/recall_engine.py

intentmind-0.2.0/src/intentmind/engines/__init__.py

@@ -1,5 +0,0 @@
-from .intent_engine import IntentEngine
-from .emotion_engine import EmotionEngine
-from .recall_engine import RecallEngine
-from .energy_engine import EnergyEngine
-__all__ = ["IntentEngine", "EmotionEngine", "RecallEngine", "EnergyEngine"]

intentmind-0.2.0/src/intentmind/engines/emotion_engine.py

@@ -1,36 +0,0 @@
-from __future__ import annotations
-
-from ..models import EmotionalState
-from ..embeddings import BaseEmbedder, cosine_similarity
-
-
-class EmotionEngine:
-    def __init__(self, embedder: BaseEmbedder):
-        self.embedder = embedder
-        self.prototypes = {
-            "nötr":    embedder.embed("bilgi istiyorum"),
-            "merak":   embedder.embed("bunu merak ediyorum keşfetmek istiyorum"),
-            "heyecan": embedder.embed("bu fikir beni heyecanlandırıyor"),
-            "güven":   embedder.embed("bu mantıklı sağlam görünüyor"),
-            "şüphe":   embedder.embed("emin değilim doğrulanması gerekiyor"),
-            "korku":   embedder.embed("riskli olabilir dikkatli olmak lazım"),
-        }
-        self.override_priority = ["korku", "şüphe", "heyecan", "merak", "güven", "nötr"]
-        self.override_threshold = 0.80
-
-    def detect(self, message_embedding) -> EmotionalState:
-        scores = {emotion: cosine_similarity(message_embedding, proto) for emotion, proto in self.prototypes.items()}
-        ranked = sorted(scores.items(), key=lambda x: x[1], reverse=True)
-        top_emotion, top_score = ranked[0]
-        second_score = ranked[1][1]
-        delta = top_score - second_score
-        for priority_emotion in self.override_priority[:2]:
-            if scores[priority_emotion] > self.override_threshold:
-                top_emotion = priority_emotion
-                top_score = scores[priority_emotion]
-                break
-        if top_score < 0.55 or delta < 0.07:
-            emotion = "nötr"
-        else:
-            emotion = top_emotion
-        return EmotionalState(current=emotion, confidence=round(top_score, 3), weak_echo=emotion in ["merak", "heyecan"])