Celes Renata
a72f336ad1
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.
97 lines
3.2 KiB
Python
97 lines
3.2 KiB
Python
"""Rejected candidate storage for audit, learning, and debugging.
|
|
|
|
Stores candidates rejected during evidence verification with structured reason codes.
|
|
Currently in-memory; designed for future persistence to v3_rejected_candidates table.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
from collections import defaultdict
|
|
|
|
from services.intelligence_pipeline_v3.verification.models import (
|
|
RejectedCandidate,
|
|
RejectionReason,
|
|
)
|
|
|
|
|
|
class RejectedCandidateStore:
|
|
"""In-memory store for rejected candidates, queryable by pipeline run and reason.
|
|
|
|
Designed to be replaced with a database-backed implementation once
|
|
the v3_rejected_candidates table is deployed. The interface is stable.
|
|
"""
|
|
|
|
def __init__(self) -> None:
|
|
self._by_run: dict[str, list[RejectedCandidate]] = defaultdict(list)
|
|
self._by_reason: dict[RejectionReason, list[RejectedCandidate]] = defaultdict(list)
|
|
self._all: list[RejectedCandidate] = []
|
|
|
|
def store(self, rejected: RejectedCandidate, run_id: str = "default") -> None:
|
|
"""Store a rejected candidate.
|
|
|
|
Args:
|
|
rejected: The rejected candidate to store.
|
|
run_id: Pipeline run identifier for grouping.
|
|
"""
|
|
self._all.append(rejected)
|
|
self._by_run[run_id].append(rejected)
|
|
self._by_reason[rejected.rejection_reason].append(rejected)
|
|
|
|
def store_batch(self, rejected_list: list[RejectedCandidate], run_id: str = "default") -> None:
|
|
"""Store multiple rejected candidates in one call.
|
|
|
|
Args:
|
|
rejected_list: List of rejected candidates to store.
|
|
run_id: Pipeline run identifier for grouping.
|
|
"""
|
|
for r in rejected_list:
|
|
self.store(r, run_id)
|
|
|
|
def get_by_pipeline_run(self, run_id: str) -> list[RejectedCandidate]:
|
|
"""Retrieve all rejected candidates for a given pipeline run.
|
|
|
|
Args:
|
|
run_id: Pipeline run identifier.
|
|
|
|
Returns:
|
|
List of rejected candidates for that run (empty if none).
|
|
"""
|
|
return list(self._by_run.get(run_id, []))
|
|
|
|
def get_by_reason(self, reason: RejectionReason) -> list[RejectedCandidate]:
|
|
"""Retrieve all rejected candidates with a specific rejection reason.
|
|
|
|
Args:
|
|
reason: The rejection reason code to filter by.
|
|
|
|
Returns:
|
|
List of rejected candidates with that reason (empty if none).
|
|
"""
|
|
return list(self._by_reason.get(reason, []))
|
|
|
|
def get_all(self) -> list[RejectedCandidate]:
|
|
"""Retrieve all stored rejected candidates.
|
|
|
|
Returns:
|
|
List of all rejected candidates.
|
|
"""
|
|
return list(self._all)
|
|
|
|
def count(self) -> int:
|
|
"""Total number of rejected candidates stored."""
|
|
return len(self._all)
|
|
|
|
def count_by_reason(self) -> dict[str, int]:
|
|
"""Count of rejected candidates grouped by reason code.
|
|
|
|
Returns:
|
|
Mapping from reason code string to count.
|
|
"""
|
|
return {reason.value: len(items) for reason, items in self._by_reason.items()}
|
|
|
|
def clear(self) -> None:
|
|
"""Remove all stored rejected candidates."""
|
|
self._by_run.clear()
|
|
self._by_reason.clear()
|
|
self._all.clear()
|