graphite-engine 0.3.1__tar.gz → 0.3.2__tar.gz

{graphite_engine-0.3.1/src/graphite_engine.egg-info → graphite_engine-0.3.2}/PKG-INFO

@@ -1,13 +1,13 @@
 Metadata-Version: 2.4
 Name: graphite-engine
-Version: 0.3.1
+Version: 0.3.2
 Summary: Open-source claim verification engine for agent-generated assertions in high-stakes domains
 Author: Min Jun Kim
 License: Apache-2.0
-Project-URL: Homepage, https://github.com/graf-research/graphite
-Project-URL: Documentation, https://github.com/graf-research/graphite#readme
-Project-URL: Repository, https://github.com/graf-research/graphite
-Project-URL: Issues, https://github.com/graf-research/graphite/issues
+Project-URL: Homepage, https://github.com/minjun1/graphite-core
+Project-URL: Documentation, https://github.com/minjun1/graphite-core#readme
+Project-URL: Repository, https://github.com/minjun1/graphite-core
+Project-URL: Issues, https://github.com/minjun1/graphite-core/issues
 Keywords: verification,claims,evidence,provenance,trust,graph
 Classifier: Development Status :: 3 - Alpha
 Classifier: Intended Audience :: Developers
@@ -111,8 +111,8 @@ Dynamic: license-file
   Or from source:
 
   ```bash
-  git clone https://github.com/graf-research/graphite.git
-  cd graphite
+  git clone https://github.com/minjun1/graphite-core.git
+  cd graphite-core
   pip install -e ".[llm]"
   export GEMINI_API_KEY="your-api-key-here"
   python examples/quickstart_verification/run.py

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/README.md

@@ -73,8 +73,8 @@
   Or from source:
 
   ```bash
-  git clone https://github.com/graf-research/graphite.git
-  cd graphite
+  git clone https://github.com/minjun1/graphite-core.git
+  cd graphite-core
   pip install -e ".[llm]"
   export GEMINI_API_KEY="your-api-key-here"
   python examples/quickstart_verification/run.py

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "graphite-engine"
-version = "0.3.1"
+version = "0.3.2"
 description = "Open-source claim verification engine for agent-generated assertions in high-stakes domains"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -37,13 +37,14 @@ dev = [
 ]
 
 [project.urls]
-Homepage = "https://github.com/graf-research/graphite"
-Documentation = "https://github.com/graf-research/graphite#readme"
-Repository = "https://github.com/graf-research/graphite"
-Issues = "https://github.com/graf-research/graphite/issues"
+Homepage = "https://github.com/minjun1/graphite-core"
+Documentation = "https://github.com/minjun1/graphite-core#readme"
+Repository = "https://github.com/minjun1/graphite-core"
+Issues = "https://github.com/minjun1/graphite-core/issues"
 
 [tool.setuptools.packages.find]
 where = ["src"]
+exclude = ["graphite._archive", "graphite._archive.*"]
 
 [tool.setuptools.package-data]
 graphite = ["py.typed"]

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/src/graphite/__init__.py

@@ -14,20 +14,40 @@ Pipeline:
 
 # ── Core schemas ──
 from .schemas import ExtractedEdge, NodeRef, Provenance, InferenceBasis, ExtractionError
-from .enums import EdgeType, NodeType, SourceType, ConfidenceLevel, AssertionMode, EvidenceType
+from .enums import (
+    EdgeType,
+    NodeType,
+    SourceType,
+    ConfidenceLevel,
+    AssertionMode,
+    EvidenceType,
+)
 from .evidence import EvidencePacket, EvidenceData
 
 # ── Trust engine primitives ──
-from .claim import Claim, ClaimType, ClaimStatus, ClaimGranularity, ReviewState, ClaimOrigin
+from .claim import (
+    Claim,
+    ClaimType,
+    ClaimStatus,
+    ClaimGranularity,
+    ReviewState,
+    ClaimOrigin,
+)
 from .claim import ConfidenceFactor, ConfidenceResult
 from .claim_store import ClaimStore
 from .confidence import ConfidenceScorer
 
 # ── Domain plugin contracts ──
-from .domain import BaseFetcher, BaseExtractor, BasePipeline, DocumentContext, DomainSpec
+from .domain import (
+    BaseFetcher,
+    BaseExtractor,
+    BasePipeline,
+    DocumentContext,
+    DomainSpec,
+)
 from .domain import register_domain, get_domain, list_domains
 
 # ── Rules ──
 from .rules import BaseRuleEngine, RuleResult, ScoreBreakdown
 
-__version__ = "0.3.1"
+__version__ = "0.3.2"

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/src/graphite/adapters/alphaearth.py

@@ -12,6 +12,7 @@ AlphaEarth Foundations:
   - Earth Engine dataset: GOOGLE/SATELLITE_EMBEDDING/V1/ANNUAL
   - GCS bucket: gs://alphaearth_foundations (Requester Pays)
 """
+
 import json
 import os
 from pathlib import Path
@@ -122,7 +123,9 @@ class AlphaEarthAdapter:
         Returns:
             numpy array of shape (64,)
         """
-        cache_key = node_id or f"bbox_{bbox[0]:.4f}_{bbox[1]:.4f}_{bbox[2]:.4f}_{bbox[3]:.4f}"
+        cache_key = (
+            node_id or f"bbox_{bbox[0]:.4f}_{bbox[1]:.4f}_{bbox[2]:.4f}_{bbox[3]:.4f}"
+        )
 
         cached = self._read_cache(cache_key, year)
         if cached is not None:

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/src/graphite/adapters/weathernext.py

@@ -12,6 +12,7 @@ WeatherNext 2:
   - Access: EE/BigQuery (requires data request form)
   - Note: Experimental dataset, not validated for real-world use
 """
+
 import json
 from pathlib import Path
 from typing import Any, Dict, List, Optional

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/src/graphite/cache.py

@@ -4,6 +4,7 @@ graphite/cache.py — Extraction result caching.
 5-part cache key prevents pollution when anything changes:
   (source_id, content_hash, extractor_version, prompt_version, model_name)
 """
+
 import hashlib
 import json
 import os

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/src/graphite/claim.py

@@ -10,6 +10,7 @@ dependency links.
 Pipeline:
   Extractor → Claim[] → (optional) Edge materialization → Graph
 """
+
 import hashlib
 from datetime import datetime, timezone
 from enum import Enum
@@ -25,25 +26,29 @@ from .schemas import Provenance
 # Claim Enums
 # ═══════════════════════════════════════
 
+
 class ClaimType(str, Enum):
     """What kind of assertion this claim makes."""
-    RELATIONSHIP = "RELATIONSHIP"       # "A supplies to B"
-    ATTRIBUTE = "ATTRIBUTE"             # "A's revenue is $10B"
-    RISK_ASSERTION = "RISK_ASSERTION"   # "A is exposed to China refining risk"
-    DEPENDENCY = "DEPENDENCY"           # "A depends on B for sole-source component"
+
+    RELATIONSHIP = "RELATIONSHIP"  # "A supplies to B"
+    ATTRIBUTE = "ATTRIBUTE"  # "A's revenue is $10B"
+    RISK_ASSERTION = "RISK_ASSERTION"  # "A is exposed to China refining risk"
+    DEPENDENCY = "DEPENDENCY"  # "A depends on B for sole-source component"
 
 
 class ClaimStatus(str, Enum):
     """Trust verdict for a claim — typically computed, not manually set."""
-    SUPPORTED = "SUPPORTED"             # Evidence meets threshold
-    WEAK = "WEAK"                       # Some evidence but below threshold
-    MIXED = "MIXED"                     # Supporting AND weakening evidence present
-    UNSUPPORTED = "UNSUPPORTED"         # No or rejected evidence
-    PENDING_REVIEW = "PENDING_REVIEW"   # Awaiting analyst decision
+
+    SUPPORTED = "SUPPORTED"  # Evidence meets threshold
+    WEAK = "WEAK"  # Some evidence but below threshold
+    MIXED = "MIXED"  # Supporting AND weakening evidence present
+    UNSUPPORTED = "UNSUPPORTED"  # No or rejected evidence
+    PENDING_REVIEW = "PENDING_REVIEW"  # Awaiting analyst decision
 
 
 class ReviewState(str, Enum):
     """Analyst review workflow state."""
+
     UNREVIEWED = "UNREVIEWED"
     APPROVED = "APPROVED"
     REJECTED = "REJECTED"
@@ -52,11 +57,12 @@ class ReviewState(str, Enum):
 
 class ClaimOrigin(str, Enum):
     """How this claim was created."""
-    EXTRACTOR = "EXTRACTOR"             # Produced by an extraction pipeline
-    AGENT = "AGENT"                     # Generated by an AI agent
-    RULE_ENGINE = "RULE_ENGINE"         # Inferred by scoring rules
-    ANALYST = "ANALYST"                 # Manually entered by a human
-    IMPORTED = "IMPORTED"               # Bulk-imported from external system
+
+    EXTRACTOR = "EXTRACTOR"  # Produced by an extraction pipeline
+    AGENT = "AGENT"  # Generated by an AI agent
+    RULE_ENGINE = "RULE_ENGINE"  # Inferred by scoring rules
+    ANALYST = "ANALYST"  # Manually entered by a human
+    IMPORTED = "IMPORTED"  # Bulk-imported from external system
 
 
 class ClaimGranularity(str, Enum):
@@ -66,6 +72,7 @@ class ClaimGranularity(str, Enum):
     SYNTHESIZED: combined from multiple atomic claims.
     THESIS: high-level investment/risk thesis for UI display.
     """
+
     ATOMIC = "ATOMIC"
     SYNTHESIZED = "SYNTHESIZED"
     THESIS = "THESIS"
@@ -75,6 +82,7 @@ class ClaimGranularity(str, Enum):
 # Confidence (explainable)
 # ═══════════════════════════════════════
 
+
 class ConfidenceFactor(BaseModel):
     """One factor contributing to the overall confidence score.
 
@@ -87,9 +95,16 @@ class ConfidenceFactor(BaseModel):
             explanation="3 independent sources confirm this claim",
         )
     """
-    name: str = Field(description="Factor name: source_count, source_diversity, directness, recency, doc_quality, counter_penalty")
-    raw_value: str = Field(description="Human-readable raw value, e.g. '3 sources', '2024-01-15'")
-    contribution: float = Field(description="Normalized contribution to final score, e.g. +0.15 or -0.08")
+
+    name: str = Field(
+        description="Factor name: source_count, source_diversity, directness, recency, doc_quality, counter_penalty"
+    )
+    raw_value: str = Field(
+        description="Human-readable raw value, e.g. '3 sources', '2024-01-15'"
+    )
+    contribution: float = Field(
+        description="Normalized contribution to final score, e.g. +0.15 or -0.08"
+    )
     direction: str = Field(default="POSITIVE", description="POSITIVE or NEGATIVE")
     explanation: str = Field(description="Why this factor matters for this claim")
 
@@ -101,12 +116,19 @@ class ConfidenceResult(BaseModel):
     The level (HIGH/MEDIUM/LOW) is derived from the score for
     backwards compatibility with existing ConfidenceLevel enum.
     """
+
     score: float = Field(ge=0.0, le=1.0, description="Computed confidence 0.0 - 1.0")
-    level: ConfidenceLevel = Field(description="Derived from score: HIGH >= 0.7, MEDIUM >= 0.4, LOW < 0.4")
-    factors: List[ConfidenceFactor] = Field(default_factory=list, description="Explainable breakdown")
+    level: ConfidenceLevel = Field(
+        description="Derived from score: HIGH >= 0.7, MEDIUM >= 0.4, LOW < 0.4"
+    )
+    factors: List[ConfidenceFactor] = Field(
+        default_factory=list, description="Explainable breakdown"
+    )
 
     @classmethod
-    def from_score(cls, score: float, factors: List[ConfidenceFactor] = None) -> "ConfidenceResult":
+    def from_score(
+        cls, score: float, factors: List[ConfidenceFactor] = None
+    ) -> "ConfidenceResult":
         """Create a ConfidenceResult from a raw score, deriving the level."""
         score = max(0.0, min(1.0, score))
         if score >= 0.7:
@@ -122,6 +144,7 @@ class ConfidenceResult(BaseModel):
 # Claim (the core primitive)
 # ═══════════════════════════════════════
 
+
 def _make_claim_id(
     subjects: List[str],
     predicate: str,
@@ -166,9 +189,15 @@ class Claim(BaseModel):
             supporting_evidence=[...],
         )
     """
+
     # Identity — deterministic from (subjects, predicate, objects)
-    claim_id: str = Field(default="", description="Auto-computed if empty. Hash of (subjects, predicate, objects)")
-    claim_text: str = Field(description="Human-readable claim: 'NVDA supplies GPUs to TSLA'")
+    claim_id: str = Field(
+        default="",
+        description="Auto-computed if empty. Hash of (subjects, predicate, objects)",
+    )
+    claim_text: str = Field(
+        description="Human-readable claim: 'NVDA supplies GPUs to TSLA'"
+    )
     claim_type: ClaimType
     granularity: ClaimGranularity = Field(
         default=ClaimGranularity.ATOMIC,
@@ -177,7 +206,9 @@ class Claim(BaseModel):
 
     # Structured triple
     subject_entities: List[str] = Field(description="NodeRef IDs: ['company:NVDA']")
-    predicate: str = Field(description="Canonical relation: SUPPLIES_TO, DEPENDS_ON, EXPOSED_TO, etc.")
+    predicate: str = Field(
+        description="Canonical relation: SUPPLIES_TO, DEPENDS_ON, EXPOSED_TO, etc."
+    )
     object_entities: List[str] = Field(description="NodeRef IDs: ['company:TSLA']")
 
     # How this claim was asserted
@@ -321,8 +352,10 @@ class Claim(BaseModel):
         self.final_status = status
         self.override_reason = reason
         self.review_state = (
-            ReviewState.APPROVED if status == ClaimStatus.SUPPORTED
-            else ReviewState.REJECTED if status == ClaimStatus.UNSUPPORTED
+            ReviewState.APPROVED
+            if status == ClaimStatus.SUPPORTED
+            else ReviewState.REJECTED
+            if status == ClaimStatus.UNSUPPORTED
             else ReviewState.NEEDS_FOLLOWUP
         )
         self.reviewed_by = reviewer
@@ -336,7 +369,9 @@ class Claim(BaseModel):
     def is_overridden(self) -> bool:
         """True if an analyst explicitly set the status (Approved/Rejected/NeedsFollowup)."""
         return self.review_state in (
-            ReviewState.APPROVED, ReviewState.REJECTED, ReviewState.NEEDS_FOLLOWUP
+            ReviewState.APPROVED,
+            ReviewState.REJECTED,
+            ReviewState.NEEDS_FOLLOWUP,
         )
 
     @property
@@ -348,6 +383,7 @@ class Claim(BaseModel):
 # LLM-Native Verification Report Models
 # ═══════════════════════════════════════
 
+
 class VerdictEnum(str, Enum):
     SUPPORTED = "SUPPORTED"
     CONFLICTED = "CONFLICTED"
@@ -362,28 +398,45 @@ class ArgumentVerdictEnum(str, Enum):
 
 class VerdictRationale(BaseModel):
     """Structured reasoning slots for meta-evaluations and transparent auditing."""
-    contradiction_type: Optional[str] = Field(default=None, description="e.g., numeric mismatch, entity mismatch")
-    missing_evidence_reason: Optional[str] = Field(default=None, description="Why the evidence was insufficient or lacking")
-    temporal_alignment: Optional[str] = Field(default=None, description="e.g., stale evidence vs current claim timeline")
+
+    contradiction_type: Optional[str] = Field(
+        default=None, description="e.g., numeric mismatch, entity mismatch"
+    )
+    missing_evidence_reason: Optional[str] = Field(
+        default=None, description="Why the evidence was insufficient or lacking"
+    )
+    temporal_alignment: Optional[str] = Field(
+        default=None, description="e.g., stale evidence vs current claim timeline"
+    )
     text: str = Field(description="Free-form rationale from the LLM judge")
 
 
 class Verdict(BaseModel):
     """A claim-level judgment returned by the verifier pipeline."""
+
     claim_id: str
     claim_text: str
     verdict: VerdictEnum
     supporting_evidence_ids: List[str] = Field(default_factory=list)
     conflicting_evidence_ids: List[str] = Field(default_factory=list)
     rationale: VerdictRationale
-    needs_human_review: bool = Field(default=False, description="Flag for high-risk or low-confidence verdicts to route to human queues")
-    cited_span: Optional[str] = Field(default=None, description="The exact span from the evidence corpus cited to make the verdict")
+    needs_human_review: bool = Field(
+        default=False,
+        description="Flag for high-risk or low-confidence verdicts to route to human queues",
+    )
+    cited_span: Optional[str] = Field(
+        default=None,
+        description="The exact span from the evidence corpus cited to make the verdict",
+    )
     model_version: str = Field(default="")
-    timestamp: str = Field(default_factory=lambda: datetime.now(timezone.utc).isoformat())
+    timestamp: str = Field(
+        default_factory=lambda: datetime.now(timezone.utc).isoformat()
+    )
 
 
 class ArgumentVerdict(BaseModel):
     """An argument-level verification representing semantic logic jumps."""
+
     text: str = Field(description="The argument chain or conclusion being evaluated")
     verdict: ArgumentVerdictEnum
     rationale: VerdictRationale
@@ -392,6 +445,7 @@ class ArgumentVerdict(BaseModel):
 
 class VerificationReport(BaseModel):
     """Top-level review object aggregating the entire verification workflow."""
+
     document_id: str
     total_claims: int = 0
     supported_count: int = 0
@@ -399,7 +453,9 @@ class VerificationReport(BaseModel):
     insufficient_count: int = 0
     grounded_argument_count: int = 0
     conclusion_jump_count: int = 0
-    risky_claim_ids: List[str] = Field(default_factory=list, description="Claim IDs flagged for human review")
+    risky_claim_ids: List[str] = Field(
+        default_factory=list, description="Claim IDs flagged for human review"
+    )
     evidence_coverage_score: float = 0.0
     verdicts: List[Verdict] = Field(default_factory=list)
     argument_verdicts: List[ArgumentVerdict] = Field(default_factory=list)

{graphite_engine-0.3.1 → graphite_engine-0.3.2}/src/graphite/claim_store.py

@@ -13,6 +13,7 @@ Key semantics:
       * Same source_id, different quote → append
       * Different source_id → always append
 """
+
 import json
 import sqlite3
 from typing import List, Optional, Set, Tuple
@@ -38,7 +39,7 @@ class ClaimStore:
             cursor = conn.cursor()
 
             # Main claims table
-            cursor.execute('''
+            cursor.execute("""
             CREATE TABLE IF NOT EXISTS claims (
                 claim_id TEXT PRIMARY KEY,
                 claim_text TEXT,
@@ -52,11 +53,15 @@ class ClaimStore:
                 confidence_score REAL,
                 full_json TEXT
             )
-            ''')
+            """)
 
             # Index for fast querying
-            cursor.execute('CREATE INDEX IF NOT EXISTS idx_claims_predicate ON claims(predicate)')
-            cursor.execute('CREATE INDEX IF NOT EXISTS idx_claims_as_of_date ON claims(as_of_date)')
+            cursor.execute(
+                "CREATE INDEX IF NOT EXISTS idx_claims_predicate ON claims(predicate)"
+            )
+            cursor.execute(
+                "CREATE INDEX IF NOT EXISTS idx_claims_as_of_date ON claims(as_of_date)"
+            )
 
             conn.commit()
 
@@ -79,9 +84,7 @@ class ClaimStore:
           - Same source_id, different quote → append
           - Different source_id → always append
         """
-        seen: Set[Tuple[str, str]] = {
-            ClaimStore._evidence_key(ev) for ev in existing
-        }
+        seen: Set[Tuple[str, str]] = {ClaimStore._evidence_key(ev) for ev in existing}
         merged = list(existing)
         for ev in incoming:
             key = ClaimStore._evidence_key(ev)
@@ -104,10 +107,12 @@ class ClaimStore:
         if existing is not None:
             # Merge evidence from the new claim into the existing one
             existing.supporting_evidence = self._merge_evidence(
-                existing.supporting_evidence, claim.supporting_evidence,
+                existing.supporting_evidence,
+                claim.supporting_evidence,
             )
             existing.weakening_evidence = self._merge_evidence(
-                existing.weakening_evidence, claim.weakening_evidence,
+                existing.weakening_evidence,
+                claim.weakening_evidence,
             )
             # Update mutable fields from the newer claim
             existing.claim_text = claim.claim_text or existing.claim_text
@@ -145,7 +150,8 @@ class ClaimStore:
             score = claim.confidence.score if claim.confidence else 0.0
             full_json = claim.model_dump_json()
 
-            cursor.execute('''
+            cursor.execute(
+                """
             INSERT INTO claims (
                 claim_id, claim_text, claim_type, granularity,
                 subject_entities, predicate, object_entities,
@@ -157,19 +163,21 @@ class ClaimStore:
                 computed_status=excluded.computed_status,
                 confidence_score=excluded.confidence_score,
                 full_json=excluded.full_json
-            ''', (
-                claim.claim_id,
-                claim.claim_text,
-                claim.claim_type.value,
-                claim.granularity.value,
-                subjects_str,
-                claim.predicate,
-                objects_str,
-                claim.as_of_date,
-                claim.computed_status.value,
-                score,
-                full_json
-            ))
+            """,
+                (
+                    claim.claim_id,
+                    claim.claim_text,
+                    claim.claim_type.value,
+                    claim.granularity.value,
+                    subjects_str,
+                    claim.predicate,
+                    objects_str,
+                    claim.as_of_date,
+                    claim.computed_status.value,
+                    score,
+                    full_json,
+                ),
+            )
 
             conn.commit()
 
@@ -182,36 +190,40 @@ class ClaimStore:
         """Retrieve a claim by its exact ID."""
         with sqlite3.connect(self.db_path) as conn:
             cursor = conn.cursor()
-            cursor.execute('SELECT full_json FROM claims WHERE claim_id = ?', (claim_id,))
+            cursor.execute(
+                "SELECT full_json FROM claims WHERE claim_id = ?", (claim_id,)
+            )
             row = cursor.fetchone()
 
             if row:
                 return Claim.model_validate_json(row[0])
             return None
 
-    def search_claims(self,
-                      subject_contains: Optional[str] = None,
-                      object_contains: Optional[str] = None,
-                      predicate: Optional[str] = None,
-                      as_of_date: Optional[str] = None) -> List[Claim]:
+    def search_claims(
+        self,
+        subject_contains: Optional[str] = None,
+        object_contains: Optional[str] = None,
+        predicate: Optional[str] = None,
+        as_of_date: Optional[str] = None,
+    ) -> List[Claim]:
         """Search claims by components."""
-        query = 'SELECT full_json FROM claims WHERE 1=1'
+        query = "SELECT full_json FROM claims WHERE 1=1"
         params = []
 
         if subject_contains:
-            query += ' AND subject_entities LIKE ?'
-            params.append(f'%{subject_contains}%')
+            query += " AND subject_entities LIKE ?"
+            params.append(f"%{subject_contains}%")
 
         if object_contains:
-            query += ' AND object_entities LIKE ?'
-            params.append(f'%{object_contains}%')
+            query += " AND object_entities LIKE ?"
+            params.append(f"%{object_contains}%")
 
         if predicate:
-            query += ' AND predicate = ?'
+            query += " AND predicate = ?"
             params.append(predicate)
 
         if as_of_date:
-            query += ' AND as_of_date = ?'
+            query += " AND as_of_date = ?"
             params.append(as_of_date)
 
         with sqlite3.connect(self.db_path) as conn:
@@ -301,8 +313,7 @@ class ClaimStore:
 
             # Must share at least one entity
             c_entities = set(
-                e.upper().strip()
-                for e in c.subject_entities + c.object_entities
+                e.upper().strip() for e in c.subject_entities + c.object_entities
             )
             if claim_entities & c_entities:
                 conflicts.append(c)