docs(spec 040): configuration audit — Testimize rewrite, missing sections, accuracy fixes

- testimize-integration.md: full rewrite for v1.48.3 in-process NuGet flow
  (was MCP child process). New: how-it-works pipeline, abc_settings sub-table,
  single-field vs multi-field paths, debug log line reference, health check
  output update. Removed all MCP references (mcp.command, mcp.args, global
  tool install).

- configuration.md: added analysis section (categories schema + defaults),
  testimize cross-reference section, ai.fallback_strategy property,
  coverage.requirements_file + criteria_import sub-properties.

- customization.md: expanded coverage settings with requirements_file,
  criteria_import sub-object, report_orphans/broken_links/mismatches flags.

- README.md: updated Testimize callout from "install global MCP tool" to
  "bundled NuGet dependency, no separate install needed".

- PROJECT-KNOWLEDGE.md: removed testimize.mcp from example config, updated
  spec 038 note for v1.48.3 in-process flow, fixed ai.debug_log_enabled →
  debug.enabled reference.

Author: angelovstanton

PROJECT-KNOWLEDGE.md

@@ -299,8 +299,7 @@ Three-section unified coverage with distinct semantics:
   "testimize": {
     "enabled": false,
     "mode": "exploratory",
-    "strategy": "HybridArtificialBeeColony",
-    "mcp": { "command": "testimize-mcp", "args": ["--mcp"] }
+    "strategy": "HybridArtificialBeeColony"
   }
 }
 ```
@@ -309,8 +308,9 @@ Three-section unified coverage with distinct semantics:
 > - `ai.analysis_timeout_minutes` (default 2) — behavior analysis timeout
 > - `ai.generation_timeout_minutes` (default 5) — per-batch generation timeout
 > - `ai.generation_batch_size` (default 30) — tests per AI call
-> - `ai.debug_log_enabled` (default true) — writes per-call timing diagnostics
->   to `.spectra-debug.log` in the project root
+> - `debug.enabled` (default false) — writes per-call timing diagnostics
+>   to `.spectra-debug.log` in the project root. Use `--verbosity diagnostic`
+>   for one-shot override without changing config.
 >
 > Slower / reasoning models (DeepSeek-V3, large Azure deployments) typically
 > need: analysis 5–10 min, generation 15–20 min, batches of 6–10.
@@ -325,8 +325,13 @@ Three-section unified coverage with distinct semantics:
 > (`github-models`, `azure-openai`, `azure-anthropic`, `openai`, `anthropic`).
 > Legacy `github` is a soft alias; legacy `google` is rejected.
 >
-> **Spec 038**: optional `testimize` section enables external Testimize MCP
-> integration for algorithmic test data optimization (off by default).
+> **Spec 038 / v1.48.3**: optional `testimize` section enables in-process
+> Testimize integration for algorithmic test data optimization (BVA, EP,
+> pairwise, ABC). The library runs in-process as a bundled NuGet dependency
+> — no MCP child process, no separate tool installation. When enabled,
+> `TestimizeRunner` calls `TestimizeEngine.Configure(...).Generate()` once
+> per suite and embeds the result in the generation prompt as authoritative
+> facts. Off by default.
 
 ## Code Conventions
 

README.md

@@ -197,13 +197,14 @@ SPECTRA is part of the [Automate The Planet](https://www.automatetheplanet.com/)
 
 **BELLATRIX** automates test execution. **Testimize** optimizes test case selection. **SPECTRA** generates and maintains the test cases themselves — closing the loop between documentation and quality assurance.
 
-> **Optional Testimize integration**: SPECTRA can call the
-> [Testimize.MCP.Server](https://github.com/AutomateThePlanet/Testimize) global
-> tool to replace AI-approximated boundary values with mathematically optimal
-> ones (BVA / EP / pairwise / ABC). Disabled by default; enable with
-> `dotnet tool install --global Testimize.MCP.Server` and set
-> `testimize.enabled` to `true` in `spectra.config.json`. See
-> [Testimize Integration](docs/testimize-integration.md).
+> **Optional Testimize integration**: SPECTRA ships with the
+> [Testimize](https://github.com/AutomateThePlanet/Testimize) library bundled
+> as an in-process NuGet dependency (no separate installation needed). When
+> enabled, it replaces AI-approximated boundary values with mathematically
+> precise ones — exact min-1/min/max/max+1 BVA, security-focused invalid
+> patterns (XSS, SQLi), and optimized multi-field pairwise/ABC combinations.
+> Disabled by default; set `testimize.enabled` to `true` in
+> `spectra.config.json`. See [Testimize Integration](docs/testimize-integration.md).
 
 ## Documentation
 

docs/configuration.md

@@ -109,6 +109,12 @@ SPECTRA is configured via `spectra.config.json` at the repository root. Run `spe
 
 All providers are accessed through the GitHub Copilot SDK.
 
+### `ai.fallback_strategy`
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `fallback_strategy` | string | `"auto"` | How to handle provider failures. `"auto"` tries the next enabled provider in priority order. When all providers fail, generation exits with an error. |
+
 ### `ai.generation_timeout_minutes`, `ai.generation_batch_size`
 
 Per-batch tuning for `spectra ai generate`. These knobs are needed for slower
@@ -312,6 +318,36 @@ and re-run.
 | `duplicate_threshold` | double | `0.6` | Similarity threshold for duplicate detection |
 | `categories` | string[] | `["happy_path", "negative", "boundary", "integration"]` | Test categories to generate |
 
+## `analysis` — Behavior Analysis Categories
+
+Controls how the AI categorizes testable behaviors during the analysis phase
+(`spectra ai generate` runs this automatically before generation). See
+[Customization > Behavior Categories](customization.md#3-behavior-categories)
+for domain-specific examples.
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `categories` | object[] | (6 built-in) | Behavior classification taxonomy injected into the analysis prompt. |
+| `categories[].name` | string | — | Category identifier (used in tags and breakdown). |
+| `categories[].description` | string | — | One-line description injected into the AI prompt to guide classification. |
+
+Default categories:
+
+```json
+{
+  "analysis": {
+    "categories": [
+      { "name": "happy_path", "description": "Core happy-path functionality" },
+      { "name": "negative", "description": "Invalid input, unauthorized access, bad state" },
+      { "name": "edge_case", "description": "Unusual but valid scenarios" },
+      { "name": "boundary", "description": "Min/max values, exact limits, off-by-one" },
+      { "name": "error_handling", "description": "System failures, timeouts, downstream errors" },
+      { "name": "security", "description": "Auth checks, input sanitization, data exposure" }
+    ]
+  }
+}
+```
+
 ## `update` — Test Update Settings
 
 | Property | Type | Default | Description |
@@ -375,9 +411,14 @@ See [Coverage](coverage.md) for how these settings are used.
 | `file_patterns` | string[] | `["*.cs", "*.ts", "*.js", "*.py", "*.java"]` | Glob patterns for automation files (legacy) |
 | `attribute_patterns` | string[] | (framework defaults) | Regex patterns for test attributes (legacy, use `scan_patterns` instead) |
 | `test_id_pattern` | string | `"TC-\\d{3,}"` | Regex for matching test IDs |
+| `requirements_file` | string | `"docs/requirements/_requirements.yaml"` | Path to requirements YAML index for requirements coverage analysis |
 | `report_orphans` | bool | `true` | Report automation files not linked to any test |
 | `report_broken_links` | bool | `true` | Report tests referencing non-existent files |
 | `report_mismatches` | bool | `true` | Report inconsistencies between index and files |
+| `criteria_import.default_source_type` | string | `"manual"` | Default source type for imported acceptance criteria |
+| `criteria_import.auto_split` | bool | `true` | Automatically split compound criteria into atomic ones during import |
+| `criteria_import.normalize_rfc2119` | bool | `true` | Normalize RFC 2119 keywords (MUST, SHALL, SHOULD) in imported criteria |
+| `criteria_import.id_prefix` | string | `"AC"` | Prefix for auto-generated acceptance criteria IDs |
 
 ## `selections` — Saved Test Selections
 
@@ -465,6 +506,22 @@ When configured, the execution agent uses the specified Copilot Space to answer
 | `validate_on_load` | bool | `true` | Validate profile structure on load |
 | `include_in_validation` | bool | `true` | Include profile validation in `spectra validate` |
 
+## `testimize` — Testimize Integration (Optional)
+
+Disabled by default. When enabled, SPECTRA runs the Testimize engine
+in-process during test generation to produce algorithmically precise
+boundary values, equivalence partitions, and multi-field combinations.
+See [Testimize Integration](testimize-integration.md) for the full setup
+guide, how-it-works flow, and ABC tuning reference.
+
+| Property | Type | Default | Description |
+|----------|------|---------|-------------|
+| `enabled` | bool | `false` | Master switch. |
+| `mode` | string | `"exploratory"` | Maps to Testimize's `TestCaseCategory`: `"exploratory"` → All, `"valid"` → Valid only, `"validation"` → Invalid only. |
+| `strategy` | string | `"HybridArtificialBeeColony"` | Algorithm for multi-field suites: `Pairwise`, `OptimizedPairwise`, `Combinatorial`, `OptimizedCombinatorial`, `HybridArtificialBeeColony`. |
+| `settings_file` | string? | `null` | Path to a `testimizeSettings.json` with per-type equivalence classes and ABC tuning. |
+| `abc_settings` | object? | `null` | Optional ABC algorithm tuning. See [Testimize Integration > abc_settings](testimize-integration.md#abc_settings-optional-tuning). |
+
 ## `debug` — Debug Logging (Spec 040)
 
 Controls the append-only `.spectra-debug.log` diagnostic file. **Disabled by

docs/customization.md

@@ -183,6 +183,33 @@ match, and where acceptance criteria live.
 - `scan_patterns` — glob patterns for automation code files.
 - `criteria_dir` — where per-document `.criteria.yaml` files are stored.
 - `criteria_file` — path to the master criteria index.
+- `requirements_file` — path to the requirements YAML index (default:
+  `"docs/requirements/_requirements.yaml"`). Used by requirements coverage
+  analysis.
+- `report_orphans` / `report_broken_links` / `report_mismatches` — boolean
+  flags (all default `true`) controlling which coverage report sections are
+  generated. Set to `false` to suppress noisy sections in large repos.
+
+**Criteria import tuning** (`coverage.criteria_import`):
+
+```json
+{
+  "coverage": {
+    "criteria_import": {
+      "default_source_type": "manual",
+      "auto_split": true,
+      "normalize_rfc2119": true,
+      "id_prefix": "AC"
+    }
+  }
+}
+```
+
+- `auto_split` — when importing compound acceptance criteria from CSV/YAML,
+  automatically split them into atomic criteria.
+- `normalize_rfc2119` — normalize RFC 2119 keywords (MUST, SHALL, SHOULD) to
+  a consistent form during import.
+- `id_prefix` — prefix for auto-generated acceptance criteria IDs (default: `"AC"`).
 
 ---