Skip to content

VeriOps documentation

Powered by VeriOps Quality Score Protocols

The VeriOps documentation pipeline generates, validates, and publishes documentation for any content type: API references, product guides, developer portals, and internal knowledge bases. This site is a live showcase produced entirely by the pipeline.

KPI dashboard

The pipeline evaluates documentation health on every run. These metrics come from reports/kpi-wall.json, generated by the autopipeline on 2026-03-21.

Metric Value Target Status
Quality score 100% 80% Excellent
Total documents 12 -- Indexed across all content types
Stale pages 0 0 No stale pages
Documentation gaps 0 0 Full coverage
Metadata completeness 100% 100% All frontmatter fields present
Protocol drift failures 0 0 All contracts valid

Supported protocols

The VeriOps API exposes five protocol interfaces. Each protocol passes through eight pipeline stages: ingest, lint, regression, docs generation, frontmatter gate, snippet lint, test assets, and publish.

Protocol Transport Contract source Stages Status
REST HTTP/1.1 + JSON OpenAPI 3.0 (api/openapi.yaml) 8/8 PASS
GraphQL HTTP POST SDL schema (graphql.schema.graphql) 8/8 PASS
gRPC HTTP/2 + Protobuf Proto3 (veriops.proto) 8/8 PASS
AsyncAPI AMQP / Kafka AsyncAPI 2.6.0 (asyncapi.yaml) 8/8 PASS
WebSocket WSS WebSocket contract (websocket.yaml) 8/8 PASS

Quality gates enforced

Every document on this site passes through 32 automated checks before publish:

Category Count What they verify
GEO checks 8 Meta description, first paragraph length, heading hierarchy, fact density, definition patterns, heading specificity -- LLM and AI search optimization
SEO checks 14 Title length, URL depth, internal links, structured data, image alt text, bare URLs, content depth, heading keywords -- traditional search engine optimization
Style checks 6 American English, active voice, no weasel words, no contractions, second person, present tense -- consistent tone across all pages
Contract checks 4 Schema validation, regression detection, snippet lint, self-verification against endpoints -- technical accuracy against source contracts

See Quality evidence for the full check-by-check breakdown with thresholds and severity levels.

Automated detection and repair

The pipeline detects documentation drift when source contracts change and regenerates affected pages automatically.

Protocol drift detected and repaired

Before: The GraphQL schema added a new priority field to the Project type, but the GraphQL playground docs still listed only id, name, status, and createdAt.

Detection: The multi_protocol_contract stage compared contracts/graphql.schema.graphql against the generated docs and flagged the missing field as a regression.

Autofix: The pipeline regenerated the GraphQL reference page, added the priority field to the schema explorer table and query examples, and re-ran all quality checks.

Result: Re-validation passed. No manual editing required.

RAG retrieval pipeline

The pipeline generates a knowledge retrieval index that powers AI-driven search and support agents. Six advanced retrieval features are enabled by default.

Metric Value
Knowledge graph nodes 957
Knowledge graph edges 817
Knowledge modules 124 auto-extracted
Retrieval precision@3 0.58
Retrieval recall@3 0.93
Hallucination rate 0.0 (all retrieved documents exist in corpus)
Advanced feature Status
Token-aware chunking (750 tokens, 100 overlap) Enabled
Hybrid search (RRF, k=60) Enabled
HyDE query expansion (gpt-4.1-mini) Enabled
Cross-encoder reranking (ms-marco-MiniLM-L-6-v2) Enabled
Embedding cache (TTL: 3,600 seconds, max: 512) Enabled
Multi-mode eval (token/semantic/hybrid/hybrid+rerank) Available

What this demo proves

  1. One pipeline, any documentation type. A single docs-ops pipeline generates and validates API references, product guides, developer tutorials, troubleshooting pages, and architecture overviews from source contracts and content.

  2. Five API protocols, one workflow. REST, GraphQL, gRPC, AsyncAPI, and WebSocket documentation is generated, validated, and published through the same eight-stage pipeline.

  3. Automated quality enforcement. 32 SEO/GEO/style/contract checks run on every page before publish. No manual review of formatting or metadata.

  4. Advanced RAG pipeline. A knowledge graph with 957 nodes and 817 edges powers AI search agents. Precision@3 reaches 0.58 and recall@3 reaches 0.93 with zero hallucination. Six retrieval features (chunking, hybrid search, HyDE, reranking, embedding cache, multi-mode eval) maximize retrieval quality.

  5. Real pipeline data. Every metric on this site comes from actual pipeline reports, not hardcoded values. Run the pipeline again and the numbers update.