---
title: "Streaming scoring and validation evidence bundles"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Streaming scoring and validation evidence bundles}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>", eval = FALSE)
library(eyeprocess)
```

## Partial and streaming scoring

```{r}
partial <- score_partial_response_pattern(
  calibrated_mirt_model,
  response_pattern = c(1, 0, 1, NA, NA, NA),
  method = "MAP"
)

stream <- score_response_stream(
  calibrated_mirt_model,
  response_pattern = c(1, 0, 1, 1, 0, 1),
  method = "MAP"
)
streaming_score_history(stream)
plot(stream)
```

Streaming scoring is an operational building block. High-stakes deployment still requires calibrated item banks, latency/stopping validation, privacy governance, and score-use rules.

## Unified evidence bundles

```{r}
bundle <- collect_validation_evidence(
  model_spec = spec,
  recovery = recovery_summary,
  coverage = coverage_audit,
  convergence = convergence_audit,
  ppc = ppc,
  stress_tests = stress_tests,
  external_validation = external_results,
  process_ablation = ablation,
  negative_controls = negative_controls,
  preflight = preflight,
  drift = drift,
  model_name = "joint_process_model"
)

validation_bundle_manifest(bundle)
cat(validation_report(bundle), sep = "\n")
plot(bundle)
export_validation_bundle(bundle, "validation-export")
```