feat: Intelligence Pipeline v3 — full implementation

Multi-stage evidence-grounded inference architecture replacing the
monolithic 9B model extraction pipeline. CPU-first specialist services
handle routine extraction while the 9B vLLM model is preserved for
semantic adjudication of ambiguous cases.

Key components:
- Capability-aware inference gateway (OpenAI-compatible + Ollama)
- Endpoint registry with DB migrations and REST API
- Sentence-aware document segmenter (property tests)
- Deterministic financial parsing with offset integrity
- Symbol resolution with ambiguity detection
- Specialist service (GLiNER2, dynamic batching, K8s deployment)
- Company-specific sentiment (FinBERT, calibration)
- Retrieval-based novelty and duplicate detection
- Confidence calibration pipeline
- Deterministic routing engine (property tests)
- 9B adjudication layer with VRAM gating
- Stock-specific impact model (features, labels, baseline, trained)
- Pipeline orchestrator (state machine, queues, leases, feature flags)
- Bounded parallelism (async workers, semaphore, load shedding)
- Observability (tracing, metrics, alerts)
- Compatibility adapter (v3→v2 golden mapping tests)
- Shadow/canary promotion framework
- Active learning and fine-tuning pipeline

Test results: 1,161 tests pass, ruff lint clean.
All 282 spec tasks completed.

services/intelligence_pipeline_v3/resolution/models.py

@@ -0,0 +1,80 @@
+"""Data models for symbol resolution results."""
+
+from __future__ import annotations
+
+from enum import Enum
+
+from pydantic import BaseModel, Field
+
+
+class MatchType(str, Enum):
+    """How a resolution candidate was matched."""
+
+    exact_ticker = "exact_ticker"
+    exact_name = "exact_name"
+    alias = "alias"
+    fuzzy = "fuzzy"
+
+
+class MentionType(str, Enum):
+    """Whether a mention is explicitly stated or inferred from context."""
+
+    explicit = "explicit"
+    inferred = "inferred"
+
+
+class UnresolvedReason(str, Enum):
+    """Why a mention could not be resolved."""
+
+    not_in_registry = "not_in_registry"
+    ambiguous = "ambiguous"
+    context_needed = "context_needed"
+
+
+class ResolutionCandidate(BaseModel):
+    """A single candidate match from the symbol registry.
+
+    Candidates are ranked by confidence, with match_type indicating how
+    the match was derived.
+    """
+
+    company_id: str = Field(description="UUID of the matched company")
+    ticker: str = Field(description="Ticker symbol of the matched company")
+    name: str = Field(description="Legal or display name of the matched company")
+    confidence: float = Field(ge=0.0, le=1.0, description="Match confidence score")
+    match_type: MatchType = Field(description="How the match was derived")
+
+
+class ResolutionResult(BaseModel):
+    """Full resolution output for a single mention.
+
+    Contains ranked candidates, ambiguity margin, and mention classification.
+    """
+
+    candidates: list[ResolutionCandidate] = Field(default_factory=list)
+    ambiguity_margin: float = Field(
+        default=1.0,
+        ge=0.0,
+        le=1.0,
+        description="Difference between top-2 candidate confidences. 1.0 = unambiguous single match, 0.0 = tied.",
+    )
+    is_ambiguous: bool = Field(
+        default=False,
+        description="True when top candidates are too close to distinguish without context.",
+    )
+    mention_type: MentionType = Field(
+        default=MentionType.explicit,
+        description="Whether the mention is explicit text or inferred exposure.",
+    )
+
+
+class UnresolvedMention(BaseModel):
+    """A mention that could not be resolved to any company in the registry.
+
+    Preserved as-is rather than having a ticker invented for it.
+    """
+
+    literal_text: str = Field(description="Original text as it appeared in the document")
+    start_char: int = Field(ge=0, description="Start character offset in source document")
+    end_char: int = Field(gt=0, description="End character offset (exclusive)")
+    reason: UnresolvedReason = Field(description="Why resolution failed")