hindsight-api 0.1.5__py3-none-any.whl → 0.1.7__py3-none-any.whl

hindsight_api/engine/search/think_utils.py

@@ -2,41 +2,35 @@
 Think operation utilities for formulating answers based on agent and world facts.
 """
 
-import asyncio
 import logging
 import re
-from datetime import datetime, timezone
-from typing import Dict, List, Any
+from datetime import datetime
+
 from pydantic import BaseModel, Field
 
-from ..response_models import ReflectResult, MemoryFact, DispositionTraits
+from ..response_models import DispositionTraits, MemoryFact
 
 logger = logging.getLogger(__name__)
 
 
 class Opinion(BaseModel):
     """An opinion formed by the bank."""
+
     opinion: str = Field(description="The opinion or perspective with reasoning included")
     confidence: float = Field(description="Confidence score for this opinion (0.0 to 1.0, where 1.0 is very confident)")
 
 
 class OpinionExtractionResponse(BaseModel):
     """Response containing extracted opinions."""
-    opinions: List[Opinion] = Field(
-        default_factory=list,
-        description="List of opinions formed with their supporting reasons and confidence scores"
+
+    opinions: list[Opinion] = Field(
+        default_factory=list, description="List of opinions formed with their supporting reasons and confidence scores"
     )
 
 
 def describe_trait_level(value: int) -> str:
     """Convert trait value (1-5) to descriptive text."""
-    levels = {
-        1: "very low",
-        2: "low",
-        3: "moderate",
-        4: "high",
-        5: "very high"
-    }
+    levels = {1: "very low", 2: "low", 3: "moderate", 4: "high", 5: "very high"}
     return levels.get(value, "moderate")
 
 
@@ -47,7 +41,7 @@ def build_disposition_description(disposition: DispositionTraits) -> str:
         2: "You tend to trust information but may question obvious inconsistencies.",
         3: "You have a balanced approach to information, neither too trusting nor too skeptical.",
         4: "You are somewhat skeptical and often question the reliability of information.",
-        5: "You are highly skeptical and critically examine all information for accuracy and hidden motives."
+        5: "You are highly skeptical and critically examine all information for accuracy and hidden motives.",
     }
 
     literalism_desc = {
@@ -55,7 +49,7 @@ def build_disposition_description(disposition: DispositionTraits) -> str:
         2: "You tend to consider context and implied meaning alongside literal statements.",
         3: "You balance literal interpretation with contextual understanding.",
         4: "You prefer to interpret information more literally and precisely.",
-        5: "You interpret information very literally and focus on exact wording and commitments."
+        5: "You interpret information very literally and focus on exact wording and commitments.",
     }
 
     empathy_desc = {
@@ -63,7 +57,7 @@ def build_disposition_description(disposition: DispositionTraits) -> str:
         2: "You consider facts first but acknowledge emotional factors exist.",
         3: "You balance factual analysis with emotional understanding.",
         4: "You give significant weight to emotional context and human factors.",
-        5: "You strongly consider the emotional state and circumstances of others when forming memories."
+        5: "You strongly consider the emotional state and circumstances of others when forming memories.",
     }
 
     return f"""Your disposition traits:
@@ -72,7 +66,7 @@ def build_disposition_description(disposition: DispositionTraits) -> str:
 - Empathy ({describe_trait_level(disposition.empathy)}): {empathy_desc.get(disposition.empathy, empathy_desc[3])}"""
 
 
-def format_facts_for_prompt(facts: List[MemoryFact]) -> str:
+def format_facts_for_prompt(facts: list[MemoryFact]) -> str:
     """Format facts as JSON for LLM prompt."""
     import json
 
@@ -80,9 +74,7 @@ def format_facts_for_prompt(facts: List[MemoryFact]) -> str:
         return "[]"
     formatted = []
     for fact in facts:
-        fact_obj = {
-            "text": fact.text
-        }
+        fact_obj = {"text": fact.text}
 
         # Add context if available
         if fact.context:
@@ -94,7 +86,7 @@ def format_facts_for_prompt(facts: List[MemoryFact]) -> str:
             if isinstance(occurred_start, str):
                 fact_obj["occurred_start"] = occurred_start
             elif isinstance(occurred_start, datetime):
-                fact_obj["occurred_start"] = occurred_start.strftime('%Y-%m-%d %H:%M:%S')
+                fact_obj["occurred_start"] = occurred_start.strftime("%Y-%m-%d %H:%M:%S")
 
         formatted.append(fact_obj)
 
@@ -176,16 +168,14 @@ def get_system_message(disposition: DispositionTraits) -> str:
     elif disposition.empathy <= 2:
         instructions.append("Focus on facts and outcomes rather than emotional context.")
 
-    disposition_instruction = " ".join(instructions) if instructions else "Balance your disposition traits when interpreting information."
+    disposition_instruction = (
+        " ".join(instructions) if instructions else "Balance your disposition traits when interpreting information."
+    )
 
     return f"You are a person with your own thoughts, experiences, opinions, and disposition. When asked a question, respond naturally like a human would - share what you genuinely think based on what you know and have experienced. {disposition_instruction} Be direct, express your views confidently, and use 'I think', 'I believe', 'in my view', etc. Respond in plain text without markdown formatting."
 
 
-async def extract_opinions_from_text(
-    llm_config,
-    text: str,
-    query: str
-) -> List[Opinion]:
+async def extract_opinions_from_text(llm_config, text: str, query: str) -> list[Opinion]:
     """
     Extract opinions with reasons and confidence from text using LLM.
 
@@ -238,11 +228,14 @@ If no genuine opinions are expressed (e.g., the response just says "I don't know
     try:
         result = await llm_config.call(
             messages=[
-                {"role": "system", "content": "You are converting opinions from text into first-person statements. Always use 'I think', 'I believe', 'I feel', etc. NEVER use third-person like 'The speaker' or 'They'."},
-                {"role": "user", "content": extraction_prompt}
+                {
+                    "role": "system",
+                    "content": "You are converting opinions from text into first-person statements. Always use 'I think', 'I believe', 'I feel', etc. NEVER use third-person like 'The speaker' or 'They'.",
+                },
+                {"role": "user", "content": extraction_prompt},
             ],
             response_format=OpinionExtractionResponse,
-            scope="memory_extract_opinion"
+            scope="memory_extract_opinion",
         )
 
         # Format opinions with confidence score and convert to first-person
@@ -253,14 +246,18 @@ If no genuine opinions are expressed (e.g., the response just says "I don't know
 
             # Replace common third-person patterns with first-person
             def singularize_verb(verb):
-                if verb.endswith('es'):
+                if verb.endswith("es"):
                     return verb[:-1]  # believes -> believe
-                elif verb.endswith('s'):
+                elif verb.endswith("s"):
                     return verb[:-1]  # thinks -> think
                 return verb
 
             # Pattern: "The speaker/user [verb]..." -> "I [verb]..."
-            match = re.match(r'^(The speaker|The user|They|It is believed) (believes?|thinks?|feels?|says|asserts?|considers?)(\s+that)?(.*)$', opinion_text, re.IGNORECASE)
+            match = re.match(
+                r"^(The speaker|The user|They|It is believed) (believes?|thinks?|feels?|says|asserts?|considers?)(\s+that)?(.*)$",
+                opinion_text,
+                re.IGNORECASE,
+            )
             if match:
                 verb = singularize_verb(match.group(2))
                 that_part = match.group(3) or ""  # Keep " that" if present
@@ -268,17 +265,96 @@ If no genuine opinions are expressed (e.g., the response just says "I don't know
                 opinion_text = f"I {verb}{that_part}{rest}"
 
             # If still doesn't start with first-person, prepend "I believe that "
-            first_person_starters = ["I think", "I believe", "I feel", "In my view", "I've come to believe", "Previously I"]
+            first_person_starters = [
+                "I think",
+                "I believe",
+                "I feel",
+                "In my view",
+                "I've come to believe",
+                "Previously I",
+            ]
             if not any(opinion_text.startswith(starter) for starter in first_person_starters):
                 opinion_text = "I believe that " + opinion_text[0].lower() + opinion_text[1:]
 
-            formatted_opinions.append(Opinion(
-                opinion=opinion_text,
-                confidence=op.confidence
-            ))
+            formatted_opinions.append(Opinion(opinion=opinion_text, confidence=op.confidence))
 
         return formatted_opinions
 
     except Exception as e:
         logger.warning(f"Failed to extract opinions: {str(e)}")
         return []
+
+
+async def reflect(
+    llm_config,
+    query: str,
+    experience_facts: list[str] = None,
+    world_facts: list[str] = None,
+    opinion_facts: list[str] = None,
+    name: str = "Assistant",
+    disposition: DispositionTraits = None,
+    background: str = "",
+    context: str = None,
+) -> str:
+    """
+    Standalone reflect function for generating answers based on facts.
+
+    This is a static version of the reflect operation that can be called
+    without a MemoryEngine instance, useful for testing.
+
+    Args:
+        llm_config: LLM provider instance
+        query: Question to answer
+        experience_facts: List of experience/agent fact strings
+        world_facts: List of world fact strings
+        opinion_facts: List of opinion fact strings
+        name: Name of the agent/persona
+        disposition: Disposition traits (defaults to neutral)
+        background: Background information
+        context: Additional context for the prompt
+
+    Returns:
+        Generated answer text
+    """
+    # Default disposition if not provided
+    if disposition is None:
+        disposition = DispositionTraits(skepticism=3, literalism=3, empathy=3)
+
+    # Convert string lists to MemoryFact format for formatting
+    def to_memory_facts(facts: list[str], fact_type: str) -> list[MemoryFact]:
+        if not facts:
+            return []
+        return [MemoryFact(id=f"test-{i}", text=f, fact_type=fact_type) for i, f in enumerate(facts)]
+
+    agent_results = to_memory_facts(experience_facts or [], "experience")
+    world_results = to_memory_facts(world_facts or [], "world")
+    opinion_results = to_memory_facts(opinion_facts or [], "opinion")
+
+    # Format facts for prompt
+    agent_facts_text = format_facts_for_prompt(agent_results)
+    world_facts_text = format_facts_for_prompt(world_results)
+    opinion_facts_text = format_facts_for_prompt(opinion_results)
+
+    # Build prompt
+    prompt = build_think_prompt(
+        agent_facts_text=agent_facts_text,
+        world_facts_text=world_facts_text,
+        opinion_facts_text=opinion_facts_text,
+        query=query,
+        name=name,
+        disposition=disposition,
+        background=background,
+        context=context,
+    )
+
+    system_message = get_system_message(disposition)
+
+    # Call LLM
+    answer_text = await llm_config.call(
+        messages=[{"role": "system", "content": system_message}, {"role": "user", "content": prompt}],
+        scope="memory_think",
+        temperature=0.9,
+        max_completion_tokens=1000,
+    )
+
+    return answer_text.strip()

hindsight_api/engine/search/trace.py

@@ -4,15 +4,18 @@ Search trace models for debugging and visualization.
 These Pydantic models define the structure of search traces, capturing
 every step of the spreading activation search process for analysis.
 """
+
 from datetime import datetime
-from typing import List, Optional, Dict, Any, Literal
+from typing import Any, Literal
+
 from pydantic import BaseModel, Field
 
 
 class QueryInfo(BaseModel):
     """Information about the search query."""
+
     query_text: str = Field(description="Original query text")
-    query_embedding: List[float] = Field(description="Generated query embedding vector")
+    query_embedding: list[float] = Field(description="Generated query embedding vector")
     timestamp: datetime = Field(description="When the query was executed")
     budget: int = Field(description="Maximum nodes to explore")
     max_tokens: int = Field(description="Maximum tokens to return in results")
@@ -20,6 +23,7 @@ class QueryInfo(BaseModel):
 
 class EntryPoint(BaseModel):
     """An entry point node selected for search."""
+
     node_id: str = Field(description="Memory unit ID")
     text: str = Field(description="Memory unit text content")
     similarity_score: float = Field(description="Cosine similarity to query", ge=0.0, le=1.0)
@@ -28,6 +32,7 @@ class EntryPoint(BaseModel):
 
 class WeightComponents(BaseModel):
     """Breakdown of weight calculation components."""
+
     activation: float = Field(description="Activation from spreading (can exceed 1.0 through accumulation)", ge=0.0)
     semantic_similarity: float = Field(description="Semantic similarity to query", ge=0.0, le=1.0)
     recency: float = Field(description="Recency weight", ge=0.0, le=1.0)
@@ -43,99 +48,120 @@ class WeightComponents(BaseModel):
 
 class LinkInfo(BaseModel):
     """Information about a link to a neighbor."""
+
     to_node_id: str = Field(description="Target node ID")
     link_type: Literal["temporal", "semantic", "entity"] = Field(description="Type of link")
-    link_weight: float = Field(description="Weight of the link (can exceed 1.0 when aggregating multiple connections)", ge=0.0)
-    entity_id: Optional[str] = Field(default=None, description="Entity ID if link_type is 'entity'")
-    new_activation: Optional[float] = Field(default=None, description="Activation that would be passed to neighbor (None for supplementary links)")
+    link_weight: float = Field(
+        description="Weight of the link (can exceed 1.0 when aggregating multiple connections)", ge=0.0
+    )
+    entity_id: str | None = Field(default=None, description="Entity ID if link_type is 'entity'")
+    new_activation: float | None = Field(
+        default=None, description="Activation that would be passed to neighbor (None for supplementary links)"
+    )
     followed: bool = Field(description="Whether this link was followed (or pruned)")
-    prune_reason: Optional[str] = Field(default=None, description="Why link was not followed (if not followed)")
-    is_supplementary: bool = Field(default=False, description="Whether this is a supplementary link (multiple connections to same node)")
+    prune_reason: str | None = Field(default=None, description="Why link was not followed (if not followed)")
+    is_supplementary: bool = Field(
+        default=False, description="Whether this is a supplementary link (multiple connections to same node)"
+    )
 
 
 class NodeVisit(BaseModel):
     """Information about visiting a node during search."""
+
     step: int = Field(description="Step number in search (1-based)")
     node_id: str = Field(description="Memory unit ID")
     text: str = Field(description="Memory unit text content")
     context: str = Field(description="Memory unit context")
-    event_date: Optional[datetime] = Field(default=None, description="When the memory occurred")
+    event_date: datetime | None = Field(default=None, description="When the memory occurred")
     access_count: int = Field(description="Number of times accessed before this search")
 
     # How this node was reached
     is_entry_point: bool = Field(description="Whether this is an entry point")
-    parent_node_id: Optional[str] = Field(default=None, description="Node that led to this one")
-    link_type: Optional[Literal["temporal", "semantic", "entity"]] = Field(default=None, description="Type of link from parent")
-    link_weight: Optional[float] = Field(default=None, description="Weight of link from parent")
+    parent_node_id: str | None = Field(default=None, description="Node that led to this one")
+    link_type: Literal["temporal", "semantic", "entity"] | None = Field(
+        default=None, description="Type of link from parent"
+    )
+    link_weight: float | None = Field(default=None, description="Weight of link from parent")
 
     # Weights
     weights: WeightComponents = Field(description="Weight calculation breakdown")
 
     # Neighbors discovered from this node
-    neighbors_explored: List[LinkInfo] = Field(default_factory=list, description="Links explored from this node")
+    neighbors_explored: list[LinkInfo] = Field(default_factory=list, description="Links explored from this node")
 
     # Ranking
-    final_rank: Optional[int] = Field(default=None, description="Final rank in results (1-based, None if not in top-k)")
+    final_rank: int | None = Field(default=None, description="Final rank in results (1-based, None if not in top-k)")
 
 
 class PruningDecision(BaseModel):
     """Records when a node was considered but not visited."""
+
     node_id: str = Field(description="Node that was pruned")
-    reason: Literal["already_visited", "activation_too_low", "budget_exhausted"] = Field(description="Why it was pruned")
+    reason: Literal["already_visited", "activation_too_low", "budget_exhausted"] = Field(
+        description="Why it was pruned"
+    )
     activation: float = Field(description="Activation value when pruned")
     would_have_been_step: int = Field(description="What step it would have been if visited")
 
 
 class SearchPhaseMetrics(BaseModel):
     """Performance metrics for a search phase."""
+
     phase_name: str = Field(description="Name of the phase")
     duration_seconds: float = Field(description="Time taken in seconds")
-    details: Dict[str, Any] = Field(default_factory=dict, description="Additional phase-specific metrics")
+    details: dict[str, Any] = Field(default_factory=dict, description="Additional phase-specific metrics")
 
 
 class RetrievalResult(BaseModel):
     """A single result from a retrieval method."""
+
     rank: int = Field(description="Rank in this retrieval method (1-based)")
     node_id: str = Field(description="Memory unit ID")
     text: str = Field(description="Memory unit text content")
     context: str = Field(default="", description="Memory unit context")
-    event_date: Optional[datetime] = Field(default=None, description="When the memory occurred")
-    fact_type: Optional[str] = Field(default=None, description="Fact type (world, experience, opinion)")
+    event_date: datetime | None = Field(default=None, description="When the memory occurred")
+    fact_type: str | None = Field(default=None, description="Fact type (world, experience, opinion)")
     score: float = Field(description="Score from this retrieval method")
     score_name: str = Field(description="Name of the score (e.g., 'similarity', 'bm25_score', 'activation')")
 
 
 class RetrievalMethodResults(BaseModel):
     """Results from a single retrieval method."""
+
     method_name: Literal["semantic", "bm25", "graph", "temporal"] = Field(description="Name of retrieval method")
-    fact_type: Optional[str] = Field(default=None, description="Fact type this retrieval was for (world, experience, opinion)")
-    results: List[RetrievalResult] = Field(description="Retrieved results with ranks")
+    fact_type: str | None = Field(
+        default=None, description="Fact type this retrieval was for (world, experience, opinion)"
+    )
+    results: list[RetrievalResult] = Field(description="Retrieved results with ranks")
     duration_seconds: float = Field(description="Time taken for this retrieval")
-    metadata: Dict[str, Any] = Field(default_factory=dict, description="Method-specific metadata")
+    metadata: dict[str, Any] = Field(default_factory=dict, description="Method-specific metadata")
 
 
 class RRFMergeResult(BaseModel):
     """A result after RRF merging."""
+
     node_id: str = Field(description="Memory unit ID")
     text: str = Field(description="Memory unit text content")
     rrf_score: float = Field(description="Reciprocal Rank Fusion score")
-    source_ranks: Dict[str, int] = Field(description="Rank in each source that contributed (method_name -> rank)")
+    source_ranks: dict[str, int] = Field(description="Rank in each source that contributed (method_name -> rank)")
     final_rrf_rank: int = Field(description="Rank after RRF merge (1-based)")
 
 
 class RerankedResult(BaseModel):
     """A result after reranking."""
+
     node_id: str = Field(description="Memory unit ID")
     text: str = Field(description="Memory unit text content")
     rerank_score: float = Field(description="Final reranking score")
     rerank_rank: int = Field(description="Rank after reranking (1-based)")
     rrf_rank: int = Field(description="Original RRF rank before reranking")
     rank_change: int = Field(description="Change in rank (positive = moved up)")
-    score_components: Dict[str, float] = Field(default_factory=dict, description="Score breakdown")
+    score_components: dict[str, float] = Field(default_factory=dict, description="Score breakdown")
 
 
 class SearchSummary(BaseModel):
     """Summary statistics about the search."""
+
     total_nodes_visited: int = Field(description="Total nodes visited")
     total_nodes_pruned: int = Field(description="Total nodes pruned")
     entry_points_found: int = Field(description="Number of entry points")
@@ -150,33 +176,36 @@ class SearchSummary(BaseModel):
     entity_links_followed: int = Field(default=0, description="Entity links followed")
 
     # Phase timings
-    phase_metrics: List[SearchPhaseMetrics] = Field(default_factory=list, description="Metrics for each phase")
+    phase_metrics: list[SearchPhaseMetrics] = Field(default_factory=list, description="Metrics for each phase")
 
 
 class SearchTrace(BaseModel):
     """Complete trace of a search operation."""
+
     query: QueryInfo = Field(description="Query information")
 
     # New 4-way retrieval architecture
-    retrieval_results: List[RetrievalMethodResults] = Field(default_factory=list, description="Results from each retrieval method")
-    rrf_merged: List[RRFMergeResult] = Field(default_factory=list, description="Results after RRF merging")
-    reranked: List[RerankedResult] = Field(default_factory=list, description="Results after reranking")
+    retrieval_results: list[RetrievalMethodResults] = Field(
+        default_factory=list, description="Results from each retrieval method"
+    )
+    rrf_merged: list[RRFMergeResult] = Field(default_factory=list, description="Results after RRF merging")
+    reranked: list[RerankedResult] = Field(default_factory=list, description="Results after reranking")
 
     # Legacy fields (kept for backward compatibility with graph/temporal visualizations)
-    entry_points: List[EntryPoint] = Field(default_factory=list, description="Entry points selected for search (legacy)")
-    visits: List[NodeVisit] = Field(default_factory=list, description="All nodes visited during search (legacy, for graph viz)")
-    pruned: List[PruningDecision] = Field(default_factory=list, description="Nodes that were pruned (legacy)")
+    entry_points: list[EntryPoint] = Field(
+        default_factory=list, description="Entry points selected for search (legacy)"
+    )
+    visits: list[NodeVisit] = Field(
+        default_factory=list, description="All nodes visited during search (legacy, for graph viz)"
+    )
+    pruned: list[PruningDecision] = Field(default_factory=list, description="Nodes that were pruned (legacy)")
 
     summary: SearchSummary = Field(description="Summary statistics")
 
     # Final results (for comparison with visits)
-    final_results: List[Dict[str, Any]] = Field(description="Final ranked results returned to user")
+    final_results: list[dict[str, Any]] = Field(description="Final ranked results returned to user")
 
-    model_config = {
-        "json_encoders": {
-            datetime: lambda v: v.isoformat()
-        }
-    }
+    model_config = {"json_encoders": {datetime: lambda v: v.isoformat()}}
 
     def to_json(self, **kwargs) -> str:
         """Export trace as JSON string."""
@@ -186,14 +215,14 @@ class SearchTrace(BaseModel):
         """Export trace as dictionary."""
         return self.model_dump()
 
-    def get_visit_by_node_id(self, node_id: str) -> Optional[NodeVisit]:
+    def get_visit_by_node_id(self, node_id: str) -> NodeVisit | None:
         """Find a visit by node ID."""
         for visit in self.visits:
             if visit.node_id == node_id:
                 return visit
         return None
 
-    def get_search_path_to_node(self, node_id: str) -> List[NodeVisit]:
+    def get_search_path_to_node(self, node_id: str) -> list[NodeVisit]:
         """Get the path from entry point to a specific node."""
         path = []
         current_visit = self.get_visit_by_node_id(node_id)
@@ -207,10 +236,10 @@ class SearchTrace(BaseModel):
 
         return path
 
-    def get_nodes_by_link_type(self, link_type: Literal["temporal", "semantic", "entity"]) -> List[NodeVisit]:
+    def get_nodes_by_link_type(self, link_type: Literal["temporal", "semantic", "entity"]) -> list[NodeVisit]:
         """Get all nodes reached via a specific link type."""
         return [v for v in self.visits if v.link_type == link_type]
 
-    def get_entry_point_nodes(self) -> List[NodeVisit]:
+    def get_entry_point_nodes(self) -> list[NodeVisit]:
         """Get all entry point visits."""
         return [v for v in self.visits if v.is_entry_point]