Pinned Loading

1. ragbolt ragbolt Public

   ragbolt is a lightweight RAG repair layer that detects retrieval, grounding, and generation failures, applies one bounded repair, re-verifies the result, and emits a structured trace. Not a framewo…

   Python

2. llm-output-contract-check llm-output-contract-check Public

   CI gate for structured LLM outputs — schema validation, snapshot diffing, and PASS/HOLD/FAIL decisions before staging promotion.

   Python

3. llm-release-governor llm-release-governor Public

   llm-release-governor is a deterministic release gate for LLM applications that decides whether a prompt, schema, or config change can be promoted based on contract evidence, leakage checks, and aud…

   Python

4. llm-evidence-gated-generation llm-evidence-gated-generation Public

   Runtime enforcement layer for LLM outputs. Verifies claims against source evidence before emit. Not an eval tool.

   Python 1

5. llm-invoice-migration-analyzer llm-invoice-migration-analyzer Public

   llm-invoice-migration-analyzer is a local CLI for analyzing whether invoice extraction workloads can safely move to cheaper LLMs by validating extracted fields against source OCR text.** It does no…

   Python

6. embcache embcache Public

   embcache is a hardware-aware embedding cache for RAG pipelines that reuses prior query/context results before hitting expensive downstream stores or model calls.** It runs CPU-first by default, wit…

   Python 2