Skip to main content

Evaluation and Roadmap

The project validates retrieval, answer quality, citation quality, performance, OCR, and deployment behavior in stages.

Evaluation framework

DimensionMethod
Retrieval qualityRecall@k, MRR, nDCG on the Golden dataset
Answer qualityLLM-as-judge and human review against expected answers
FaithfulnessCheck every answer claim against retrieved chunks
Citation accuracyConfirm answer citations point to supporting chunks
Performancep50/p95/p99 latency, tokens/sec, concurrent request envelope
OCR qualityConfidence distribution, review rate, corrected-output acceptance

Golden dataset

The Golden dataset is the acceptance benchmark: representative documents plus labeled query-answer pairs and source-passage mappings. If passage-level labels are not available, document-level labels are the fallback with lower precision.

Milestones

StageScope
FoundationSingle-shot retrieve to synthesize, multi-format ingestion, OCR confidence gating, hybrid retrieval, grounded answers, Policy Guard, Langfuse, and upload/status APIs
Agentic expansionRouter, Planner, Executor, tool registry, multi-turn memory, session resume, and query understanding
Governance and scaleRBAC/ABAC, full audit logging, KTP extraction, human-review workflows, source connectors, and scale hardening
AcceptanceFull Golden dataset, model evaluation report, dashboard data APIs, technical documentation, and training

Validation gates

  • Service-contract validation for API behavior.
  • Integration validation for PostgreSQL/pgvector and object storage paths.
  • End-to-end validation for ingest, OCR, retrieval, and query.
  • GPU validation for LLM, embedding, OCR, and reranker.
  • Production-stack evaluation against the Golden dataset before acceptance.