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.
This commit is contained in:
Celes Renata
@@ -0,0 +1,122 @@
|
||||
-- Migration 040: Inference Registry
|
||||
-- Creates tables for the capability-aware inference gateway:
|
||||
-- inference_endpoints, model_deployments, agent_stage_bindings
|
||||
-- Adds lineage columns to agent_performance_log for v3 provenance tracking.
|
||||
|
||||
-- ─── Helper: auto-update updated_at on row modification ───────────────────────
|
||||
CREATE OR REPLACE FUNCTION update_updated_at_column()
|
||||
RETURNS TRIGGER AS $$
|
||||
BEGIN
|
||||
NEW.updated_at = now();
|
||||
RETURN NEW;
|
||||
END;
|
||||
$$ LANGUAGE plpgsql;
|
||||
|
||||
-- ─── inference_endpoints ──────────────────────────────────────────────────────
|
||||
-- Stores registered inference service endpoints (Ollama, vLLM, OpenAI-compat, specialist).
|
||||
CREATE TABLE IF NOT EXISTS inference_endpoints (
|
||||
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
|
||||
name TEXT NOT NULL UNIQUE,
|
||||
protocol TEXT NOT NULL CHECK (protocol IN ('ollama_native', 'openai_chat', 'specialist_http')),
|
||||
base_url TEXT NOT NULL,
|
||||
auth_secret_ref TEXT,
|
||||
auth_scheme TEXT NOT NULL DEFAULT 'bearer',
|
||||
default_headers JSONB NOT NULL DEFAULT '{}',
|
||||
health_path TEXT,
|
||||
enabled BOOLEAN NOT NULL DEFAULT TRUE,
|
||||
revision INTEGER NOT NULL DEFAULT 1,
|
||||
created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
||||
updated_at TIMESTAMPTZ NOT NULL DEFAULT now()
|
||||
);
|
||||
|
||||
CREATE INDEX IF NOT EXISTS idx_inference_endpoints_protocol
|
||||
ON inference_endpoints(protocol);
|
||||
|
||||
-- Auto-update updated_at on inference_endpoints changes
|
||||
DROP TRIGGER IF EXISTS trg_inference_endpoints_updated_at ON inference_endpoints;
|
||||
CREATE TRIGGER trg_inference_endpoints_updated_at
|
||||
BEFORE UPDATE ON inference_endpoints
|
||||
FOR EACH ROW EXECUTE FUNCTION update_updated_at_column();
|
||||
|
||||
-- ─── model_deployments ────────────────────────────────────────────────────────
|
||||
-- A model served by an endpoint, with declared capabilities and limits.
|
||||
CREATE TABLE IF NOT EXISTS model_deployments (
|
||||
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
|
||||
endpoint_id UUID NOT NULL REFERENCES inference_endpoints(id) ON DELETE CASCADE,
|
||||
served_model_name TEXT NOT NULL,
|
||||
display_name TEXT NOT NULL,
|
||||
capabilities JSONB NOT NULL,
|
||||
context_window INTEGER,
|
||||
max_output_tokens INTEGER,
|
||||
quantization TEXT,
|
||||
runtime_metadata JSONB NOT NULL DEFAULT '{}',
|
||||
enabled BOOLEAN NOT NULL DEFAULT TRUE,
|
||||
revision INTEGER NOT NULL DEFAULT 1,
|
||||
created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
||||
updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
||||
UNIQUE(endpoint_id, served_model_name)
|
||||
);
|
||||
|
||||
CREATE INDEX IF NOT EXISTS idx_model_deployments_endpoint
|
||||
ON model_deployments(endpoint_id);
|
||||
|
||||
-- Auto-update updated_at on model_deployments changes
|
||||
DROP TRIGGER IF EXISTS trg_model_deployments_updated_at ON model_deployments;
|
||||
CREATE TRIGGER trg_model_deployments_updated_at
|
||||
BEFORE UPDATE ON model_deployments
|
||||
FOR EACH ROW EXECUTE FUNCTION update_updated_at_column();
|
||||
|
||||
-- ─── agent_stage_bindings ─────────────────────────────────────────────────────
|
||||
-- Maps an agent + pipeline stage to one or more ordered model deployments.
|
||||
CREATE TABLE IF NOT EXISTS agent_stage_bindings (
|
||||
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
|
||||
agent_id UUID NOT NULL REFERENCES ai_agents(id) ON DELETE CASCADE,
|
||||
stage TEXT NOT NULL,
|
||||
model_deployment_id UUID REFERENCES model_deployments(id) ON DELETE SET NULL,
|
||||
route_order INTEGER NOT NULL DEFAULT 0,
|
||||
routing_config JSONB NOT NULL DEFAULT '{}',
|
||||
is_active BOOLEAN NOT NULL DEFAULT TRUE,
|
||||
revision INTEGER NOT NULL DEFAULT 1,
|
||||
created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
||||
updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
||||
UNIQUE(agent_id, stage, route_order)
|
||||
);
|
||||
|
||||
CREATE INDEX IF NOT EXISTS idx_agent_stage_bindings_agent
|
||||
ON agent_stage_bindings(agent_id);
|
||||
|
||||
CREATE INDEX IF NOT EXISTS idx_agent_stage_bindings_deployment
|
||||
ON agent_stage_bindings(model_deployment_id);
|
||||
|
||||
-- Auto-update updated_at on agent_stage_bindings changes
|
||||
DROP TRIGGER IF EXISTS trg_agent_stage_bindings_updated_at ON agent_stage_bindings;
|
||||
CREATE TRIGGER trg_agent_stage_bindings_updated_at
|
||||
BEFORE UPDATE ON agent_stage_bindings
|
||||
FOR EACH ROW EXECUTE FUNCTION update_updated_at_column();
|
||||
|
||||
-- ─── Additive lineage columns on agent_performance_log ────────────────────────
|
||||
-- Tracks which endpoint/deployment/binding was used for each logged invocation.
|
||||
-- NOTE: Revision increment logic is handled at the application layer:
|
||||
-- each UPDATE to inference_endpoints, model_deployments, or agent_stage_bindings
|
||||
-- should increment the revision column (enforced by service code, not DB trigger,
|
||||
-- to allow flexible conflict resolution).
|
||||
|
||||
ALTER TABLE agent_performance_log
|
||||
ADD COLUMN IF NOT EXISTS endpoint_id UUID REFERENCES inference_endpoints(id) ON DELETE SET NULL;
|
||||
|
||||
ALTER TABLE agent_performance_log
|
||||
ADD COLUMN IF NOT EXISTS deployment_id UUID REFERENCES model_deployments(id) ON DELETE SET NULL;
|
||||
|
||||
ALTER TABLE agent_performance_log
|
||||
ADD COLUMN IF NOT EXISTS binding_revision INTEGER;
|
||||
|
||||
ALTER TABLE agent_performance_log
|
||||
ADD COLUMN IF NOT EXISTS structured_mode TEXT;
|
||||
|
||||
CREATE INDEX IF NOT EXISTS idx_agent_perf_endpoint
|
||||
ON agent_performance_log(endpoint_id)
|
||||
WHERE endpoint_id IS NOT NULL;
|
||||
|
||||
CREATE INDEX IF NOT EXISTS idx_agent_perf_deployment
|
||||
ON agent_performance_log(deployment_id)
|
||||
WHERE deployment_id IS NOT NULL;
|
||||
Reference in New Issue
Block a user