icdev 0.0.3__py3-none-any.whl

docs/features/phase-65-writeguard.md

@@ -0,0 +1,139 @@
+# Phase 65: WriteGuard — AI Writing Assistant
+
+| Field | Value |
+|-------|-------|
+| Phase | 65 |
+| Status | Complete |
+| Date | 2026-03-04 |
+| ADRs | D-WG-1 through D-WG-14 |
+
+## Problem
+
+No FedRAMP-authorized writing assistant exists. Grammarly is cloud-only, not CUI-aware, banned by NASA, and cannot operate in air-gapped IL5/IL6 environments. Government proposal writers lack tools that address CUI marking compliance, RFP shall-statement coverage, win theme threading, cross-volume consistency, and Plain Writing Act scoring.
+
+## Solution
+
+WriteGuard is a Grammarly-like writing assistant that runs fully air-gapped via local Ollama models, integrates with ICDEV's existing RAG/LLM/marketplace infrastructure, and adds GovProposal-specific features.
+
+## Goals
+
+- General-purpose + gov-specific tone profiles
+- Fully air-gapped (Ollama only, deterministic-first)
+- Real-time inline + batch document review
+- Per-tenant style guides with ISSO-lockable rules (5-layer cascade)
+- CUI/classification-aware rules
+- Plagiarism detection (internal corpus via RAG)
+- AI-generated content detection (statistical)
+- Snippet/template management
+- GovProposal integration (shall-compliance, win themes, cross-volume consistency)
+- Marketplace asset packaging
+
+## Architecture
+
+### Two-Tier Routing
+
+| Function | Tier | Method |
+|---|---|---|
+| Grammar check | Deterministic | Regex rules |
+| Readability score | Deterministic | Flesch-Kincaid, Gunning Fog, SMOG |
+| AI content detection | Deterministic | Perplexity, burstiness, n-gram stats |
+| Tone profiler | Scanner (qwen3) | Keyword + LLM classification |
+| Coherence analyzer | Worker (qwen3 → Claude) | Cross-paragraph flow analysis |
+| Rewriter | Worker (qwen3 → Claude) | Text improvement |
+| Plagiarism check | Deterministic (RAG) | Vector similarity search |
+| Style enforcer | Deterministic | Rule evaluation against cascade |
+
+### Style Guide 5-Layer Cascade
+
+Platform → Tenant → Program → Project → User. ISSO-lockable dimensions prevent downstream overrides on compliance-critical rules.
+
+### GovProposal Bridge (D-WG-8)
+
+Read-only bridge to proposal tables. Never writes to govcon tables. Provides:
+- `check_shall_compliance()` — RFP shall-statement coverage checking
+- `validate_win_themes()` — Win theme presence in drafts
+- `check_cross_volume_consistency()` — Cross-volume numerical contradiction detection
+
+## Database Tables (6 new)
+
+| Table | Purpose |
+|-------|---------|
+| `wg_style_guides` | 5-layer cascade, version-immutable (composite PK: id, version) |
+| `wg_style_guide_locks` | ISSO-lockable dimensions |
+| `wg_analysis_results` | Append-only analysis runs (D6) |
+| `wg_analysis_findings` | Individual findings (D6) |
+| `wg_snippets` | Reusable writing templates |
+| `wg_batch_runs` | Batch analysis sessions |
+
+## Files (14 tools)
+
+| File | Purpose | LOC |
+|------|---------|-----|
+| `tools/writing/analysis_engine.py` | Core orchestrator | ~400 |
+| `tools/writing/grammar_checker.py` | Regex grammar rules | ~300 |
+| `tools/writing/readability_scorer.py` | 5 readability formulas | ~200 |
+| `tools/writing/tone_profiler.py` | Keyword + scanner | ~250 |
+| `tools/writing/style_enforcer.py` | Style guide rule evaluation | ~350 |
+| `tools/writing/coherence_analyzer.py` | Paragraph flow analysis | ~250 |
+| `tools/writing/plagiarism_detector.py` | RAG similarity wrapper | ~200 |
+| `tools/writing/ai_content_detector.py` | Statistical AI detection | ~300 |
+| `tools/writing/rewriter.py` | LLM-assisted rewrite | ~200 |
+| `tools/writing/snippet_manager.py` | CRUD + search | ~300 |
+| `tools/writing/style_guide_manager.py` | 5-layer cascade | ~450 |
+| `tools/writing/batch_analyzer.py` | Multi-doc batch mode | ~300 |
+| `tools/writing/govcon_bridge.py` | GovProposal integration | ~350 |
+| `tools/writing/signal_registrar.py` | Innovation Engine bridge | ~150 |
+
+## Configuration
+
+- `args/writeguard_config.yaml` — WriteGuard configuration
+- `context/writing/style_guides/*.yaml` — 5 default style guides (government, technical, business, academic, proposal)
+- `context/writing/grammar_rules/common_errors.json` — 30+ regex grammar patterns
+- `context/writing/grammar_rules/govcon_vocabulary.json` — FAR/DFARS/NIST terms
+
+## Dashboard
+
+- `/writeguard` — Main writing assistant page with stat grid, analyze/rewrite panel, findings display, recent analyses
+
+## Testing
+
+```bash
+pytest tests/test_writeguard_*.py -v    # 192 tests across 8 files
+```
+
+| Test File | Tests | Coverage |
+|-----------|-------|----------|
+| `test_writeguard_grammar.py` | ~30 | Grammar pattern detection |
+| `test_writeguard_readability.py` | ~20 | Readability scoring accuracy |
+| `test_writeguard_style.py` | ~35 | Style guide cascade, locks, versioning |
+| `test_writeguard_ai_detect.py` | ~25 | AI content detection |
+| `test_writeguard_plagiarism.py` | ~20 | Tone, coherence, rewriter |
+| `test_writeguard_engine.py` | ~30 | Full analysis pipeline |
+| `test_writeguard_govcon.py` | ~25 | GovProposal bridge, batch, signals |
+| `test_writeguard_snippets.py` | ~20 | Snippet CRUD |
+
+## Security
+
+- `writeguard` security gate in `args/security_gates.yaml`
+- Blocking: `cui_marking_violations_in_production`, `classification_vocabulary_leak`
+- Warning: `readability_grade_above_14`, `passive_voice_above_40_pct`, `quality_score_below_threshold`
+- Append-only tables: `wg_analysis_results`, `wg_analysis_findings`
+
+## ADRs
+
+| ADR | Decision |
+|-----|----------|
+| D-WG-1 | Independent `tools/writing/` for marketplace portability |
+| D-WG-2 | Deterministic-first pipeline (regex before LLM) |
+| D-WG-3 | 5-layer cascade with ISSO locks |
+| D-WG-4 | Inline + batch modes (same engine) |
+| D-WG-5 | Plagiarism via RAG similarity (0.85 threshold) |
+| D-WG-6 | AI detection is deterministic (advisory-only) |
+| D-WG-7 | Snippets follow knowledge_base.py pattern |
+| D-WG-8 | GovProposal via bridge, not coupling |
+| D-WG-9 | Append-only results (NIST AU compliant) |
+| D-WG-10 | 3 marketplace assets |
+| D-WG-11 | RAG source registration |
+| D-WG-12 | Two-tier routing per function |
+| D-WG-13 | CUI-aware rules engine |
+| D-WG-14 | Innovation signal registration |

docs/features/phase-66-marketplace-commerce.md

@@ -0,0 +1,79 @@
+# Phase 66: Marketplace Commerce Extension
+CUI // SP-CTI
+
+## Overview
+
+Extends Phase 37 marketplace with billing, licensing, and module runtime integration.
+Enables free and paid module subscriptions (Stripe + Gov PO), RSA-signed license tokens
+with 30-day offline grace, and a Jinja2 context processor for conditional template blocks.
+
+First integration: WriteGuard "Analyze Writing" button on GovProposal section editor,
+visible only when the writeguard module is enabled.
+
+## Architecture Decisions
+
+- **D-MKT-C1:** Separate RSA-4096 key pair for module licenses
+- **D-MKT-C2:** Reuses license_validator.py canonical JSON + RSA-SHA256 pattern
+- **D-MKT-C3:** UNIQUE(asset_id, tenant_id, user_id) at application layer
+- **D-MKT-C4:** Process-local cache, 60s TTL for `is_module_enabled()`
+- **D-MKT-C5:** marketplace_invoices is append-only (NIST AU)
+- **D-MKT-C6:** marketplace_licenses status transitions only (effectively append-only)
+- **D-MKT-C7:** Stripe is optional dependency (air-gap safe)
+- **D-MKT-C8:** Pre-coded `{% if module_enabled('slug') %}` conditionals (no XSS)
+- **D-MKT-C9:** Gov PO approval requires CO or admin role
+- **D-MKT-C10:** Free modules: no expiry, no payment, instant activation
+
+## New DB Tables (5)
+
+| Table | Purpose |
+|-------|---------|
+| marketplace_pricing | Per-asset pricing tiers (free, monthly, annual, po_only) |
+| marketplace_subscriptions | Active subscriptions with status lifecycle |
+| marketplace_invoices | Payment history (append-only) |
+| marketplace_payment_methods | Stripe customer IDs + gov PO refs |
+| marketplace_licenses | RSA-signed license tokens per subscription |
+
+## New Python Tools (6)
+
+| File | LOC | Purpose |
+|------|-----|---------|
+| tools/marketplace/license_manager.py | ~340 | RSA sign/verify, is_module_active, issue/revoke |
+| tools/marketplace/subscription_manager.py | ~300 | Create/activate/cancel subscriptions, invoicing |
+| tools/marketplace/stripe_billing.py | ~220 | Stripe Checkout, Billing Portal, webhooks |
+| tools/marketplace/gov_po_processor.py | ~230 | PO submission, CO approval workflow |
+| tools/marketplace/module_runtime.py | ~160 | Flask context processor, cache, @require_module |
+| tools/marketplace/renewal_scheduler.py | ~170 | Background phone-home, license expiry |
+
+## Dashboard Routes
+
+| Route | Method | Purpose |
+|-------|--------|---------|
+| /marketplace | GET | Browse modules, enable/subscribe |
+| /marketplace/subscriptions | GET | Manage subscriptions, view invoices |
+| /marketplace/po-queue | GET | PO approval queue (CO/admin) |
+| /api/marketplace/enable | POST | Enable free module |
+| /api/marketplace/subscribe | POST | Start paid subscription |
+| /api/marketplace/cancel | POST | Cancel subscription |
+| /api/marketplace/po/submit | POST | Submit gov purchase order |
+| /api/marketplace/po/approve | POST | CO approves PO |
+| /webhooks/stripe | POST | Stripe webhook handler |
+| /proposals/<opp>/sections/<sec> | GET | Section editor with WriteGuard |
+
+## Test Coverage
+
+- 42 unit tests in tests/test_marketplace_commerce.py
+- License signing, verification, tampering detection
+- Free and paid subscription lifecycle
+- Module runtime cache + invalidation
+- Gov PO submit/approve/reject flow
+- Grace period and expiry logic
+- Stripe unavailable handling
+- End-to-end free enable + paid PO flows
+
+## Verification
+
+1. Free enable: Click Enable -> subscription active -> license issued -> module_enabled returns True
+2. Section editor: "Analyze Writing" button visible when writeguard enabled
+3. Grace period: Token expired but within 30-day grace -> still active
+4. Air-gap: HAS_STRIPE=False -> PO flow still works
+5. All 234 tests pass (192 existing + 42 new)

docs/features/phase-67-knowledge-ingestion-rag-autodraft.md

@@ -0,0 +1,97 @@
+# CUI // SP-CTI
+# Phase 67: Knowledge Ingestion & RAG-Assisted Auto-Draft
+
+## Overview
+
+Phase 67 bridges the gap between document extraction and knowledge base storage,
+and adds RAG-assisted auto-draft generation for proposal sections. Old RFP
+responses (PDF, DOCX, TXT) can now be uploaded, chunked, classified, and stored
+as reusable knowledge blocks. When drafting a new proposal section, the system
+searches the knowledge base, retrieves relevant blocks, and generates a draft
+using the LLM router (with fine-tuned model support via D-FT-6).
+
+## Architecture Decisions
+
+| ID | Decision | Rationale |
+|----|----------|-----------|
+| D-KB-1 | Chunk by section headers first, then paragraphs if >2000 chars | Mirrors RFP structure |
+| D-KB-2 | Auto-categorize via keyword patterns (no LLM) | Air-gap safe, GOTCHA-compliant |
+| D-KB-3 | Four-section prompt template | Separation of concerns |
+| D-KB-4 | Route via `router.invoke("proposal_drafting")` | Reuses existing FT override + two-tier |
+| D-KB-5 | Flask multipart upload, 50MB max, UUID prefix | Standard pattern, prevents collisions |
+| D-KB-6 | Content hash dedup on ingestion | Prevents duplicate blocks |
+| D-KB-7 | `_read_file_content()` used as pure function | Avoids cross-DB dependency |
+
+## New Files
+
+| File | Purpose |
+|------|---------|
+| `tools/govcon/knowledge_ingestion.py` | Document → chunk → classify → knowledge_base.add_block() |
+| `tools/govcon/draft_orchestrator.py` | RAG search → prompt → LLM → save draft |
+| `hardprompts/proposal_draft.md` | System prompt template for draft generation |
+| `tools/dashboard/templates/knowledge/index.html` | Knowledge base browse, search, upload UI |
+
+## Modified Files
+
+| File | Changes |
+|------|---------|
+| `tools/finetune/pair_generator.py` | Bug fix: `draft_text` → `draft_content` |
+| `tools/rag/source_registry.py` | Bug fix: `draft_text` → `draft_content` |
+| `tools/db/init_icdev_db.py` | ALTER TABLE: add provenance columns |
+| `tools/dashboard/app.py` | 6 new routes (knowledge browse/upload/search/stats, generate-draft, draft-status) |
+| `tools/dashboard/templates/govcon.html` | Knowledge Base button |
+| `tools/dashboard/templates/proposals/section_detail.html` | Generate Draft, Save, approval workflow |
+| `tools/dashboard/templates/proposals/detail.html` | Generate action in section rows |
+
+## Data Flow
+
+```
+INGESTION:
+  Upload PDF/DOCX → Flask API → knowledge_ingestion.ingest_document()
+    → document_extractor._read_file_content() → _chunk_document()
+    → _auto_categorize() → knowledge_base.add_block() per chunk
+
+AUTO-DRAFT:
+  Click "Generate Draft" → draft_orchestrator.generate_draft()
+    → Load section + opportunity + shall-statements
+    → knowledge_base.search_blocks() (hybrid keyword + RAG semantic)
+    → Fill hardprompts/proposal_draft.md template
+    → router.invoke("proposal_drafting") → save to proposal_section_drafts
+
+FEEDBACK LOOP:
+  Approve draft → pair_generator --generate-from-govcon → fine-tuning pairs
+    → Fine-tuned model → better future drafts
+```
+
+## Dashboard Routes
+
+| Route | Method | Purpose |
+|-------|--------|---------|
+| `/govcon/knowledge-base` | GET | Browse/search knowledge blocks |
+| `/api/govcon/knowledge-base/upload` | POST | File upload → ingest_document() |
+| `/api/govcon/knowledge-base/search` | GET | Search blocks |
+| `/api/govcon/knowledge-base/stats` | GET | Stats (total, by category/domain) |
+| `/api/govcon/sections/<id>/generate-draft` | POST | Auto-draft generation |
+| `/api/govcon/sections/<id>/draft-status` | POST | Approval workflow |
+
+## CLI Usage
+
+```bash
+# Ingest a single document
+python tools/govcon/knowledge_ingestion.py --ingest --file old_rfp.pdf --json
+
+# Bulk ingest a directory
+python tools/govcon/knowledge_ingestion.py --bulk-ingest --directory ./old_rfps --json
+
+# Generate draft for a section
+python tools/govcon/draft_orchestrator.py --generate --section-id <id> --opportunity-id <id> --json
+
+# Preview context without LLM
+python tools/govcon/draft_orchestrator.py --preview --section-id <id> --opportunity-id <id> --json
+```
+
+## Bug Fixes
+
+- `pair_generator.py`: `draft_text` → `draft_content`, `quality_score` → `confidence_score`, `updated_at` → `created_at`
+- `source_registry.py`: `draft_text` → `draft_content` in content_cols and metadata_cols
+- `draft_orchestrator.py`: `shall_text` → `statement_text`, `confidence_score` → `confidence`, `draft_method` column removed (doesn't exist)

docs/features/phase-68-enhanced-autodraft-pipeline.md

@@ -0,0 +1,109 @@
+# CUI // SP-CTI
+# Phase 68: Enhanced Auto-Draft Pipeline
+
+## Overview
+
+Phase 68 upgrades the Phase 67 single-step RAG draft generation into a 4-step
+enhanced pipeline: KB Retrieval → Engine Enrichment → SMART Solutioning → Content
+Generation with WriteGuard quality gate. Adds batch "Generate All" with
+dependency-aware ordering and step-by-step progress UI.
+
+## Architecture Decisions
+
+| ID | Decision | Rationale |
+|----|----------|-----------|
+| D-P68-1 | New `engine_enrichment.py` for DB keyword queries | GOTCHA separation; standalone for Phase 69 extraction |
+| D-P68-2 | Static `context/icdev_methodology.md` + KB search | Air-gap safe; deterministic context injection |
+| D-P68-3 | New `draft_generation_jobs` table for batch tracking | Polling-based progress needs persistent state |
+| D-P68-4 | WriteGuard post-check via direct `analyze()` call | Same-process, faster than HTTP roundtrip |
+| D-P68-5 | Auto-rewrite once, then flag human | Balance automation with HITL quality control |
+| D-P68-6 | Kahn's topological sort for batch ordering | Deterministic, handles deps + sort_order fallback |
+| D-P68-7 | Background thread for batch generation | Long-running, UI polls for progress |
+| D-P68-8 | Prior sections context injected into each subsequent section | Sequential dependency builds coherent proposal |
+
+## New Files
+
+| File | Purpose |
+|------|---------|
+| `tools/govcon/engine_enrichment.py` | DB queries against Innovation/Creative/Research tables |
+| `context/icdev_methodology.md` | Static ICDEV framework reference (GOTCHA, ATLAS, RICOAS, ZTA) |
+
+## Modified Files
+
+| File | Changes |
+|------|---------|
+| `tools/govcon/draft_orchestrator.py` | Added v2 pipeline, batch generation, job tracking, SMART plan |
+| `hardprompts/proposal_draft.md` | Added engine_enrichments, smart_plan, prior_sections, style_hints |
+| `tools/dashboard/app.py` | Updated generate-draft to v2, added generate-all + job progress routes |
+| `tools/dashboard/templates/proposals/detail.html` | Generate All button, progress bar, polling JS |
+| `tools/dashboard/templates/proposals/section_detail.html` | Pipeline steps panel, quality score display |
+| `tools/db/init_icdev_db.py` | `draft_generation_jobs` table, ALTER columns on `proposal_section_drafts` |
+
+## Pipeline Data Flow
+
+```
+Step 1: KB Retrieval
+  → knowledge_base.search_blocks(query, domain, top_k=5) → kb_blocks[]
+  → search_blocks("GOTCHA ATLAS methodology", top_k=3) → methodology_blocks[]
+
+Step 2: Engine Enrichment
+  → engine_enrichment.extract_rfp_keywords(section, opp, shalls) → keywords[]
+  → query_innovation_tables(keywords) → innovation_results[]
+  → query_creative_tables(keywords) → creative_results[]
+  → query_research_tables(keywords) → research_results[]
+
+Step 3: SMART Solutioning
+  → Read context/icdev_methodology.md (static, ~3000 chars max)
+  → Build SMART prompt combining KB + engines + methodology
+  → router.invoke("proposal_drafting") → structured SMART plan
+  → Fallback: deterministic bullet-point plan from KB blocks
+
+Step 4: Content Generation + Quality Gate
+  → Fill enhanced hardprompts/proposal_draft.md with ALL context
+  → router.invoke("proposal_drafting") → draft_content
+  → WriteGuard analyze(draft_content) → quality_score
+  → IF score < 60: rewriter.rewrite() → rewritten draft
+  → IF still < 60: flagged for human review
+  → Save to proposal_section_drafts with provenance
+```
+
+## Dashboard Routes
+
+| Route | Method | Purpose |
+|-------|--------|---------|
+| `/api/govcon/sections/<id>/generate-draft` | POST | Single section v2 pipeline (updated) |
+| `/api/govcon/proposals/<opp_id>/generate-all` | POST | Start batch job, returns job_id |
+| `/api/govcon/jobs/<job_id>/progress` | GET | Poll batch progress |
+
+## Schema Changes
+
+### New Table: `draft_generation_jobs`
+Tracks batch generation jobs with step-by-step progress.
+
+### New Columns on `proposal_section_drafts`
+- `engine_enrichment_ids` — JSON array of engine result IDs used
+- `smart_plan` — Generated SMART plan text
+- `writeguard_score` — Quality score from WriteGuard post-check
+- `writeguard_result_id` — Reference to WriteGuard analysis result
+- `was_auto_rewritten` — Whether auto-rewrite was applied
+- `pipeline_version` — "v1" or "v2"
+
+## CLI Usage
+
+```bash
+# Single section v2 pipeline
+python tools/govcon/draft_orchestrator.py --generate-v2 --section-id <id> --opportunity-id <id> --json
+
+# Batch generate all sections
+python tools/govcon/draft_orchestrator.py --batch --opportunity-id <id> --json
+
+# Engine enrichment test
+python tools/govcon/engine_enrichment.py --test --json
+
+# v1 pipeline (legacy fallback)
+python tools/govcon/draft_orchestrator.py --generate --section-id <id> --opportunity-id <id> --json
+```
+
+## Bug Fixes
+
+- Applied Phase 68 schema to both `sparkpilot.db` and `icdev.db` (dashboard uses `icdev.db`)

docs/features/phase-69-proposalai-marketplace-module.md

@@ -0,0 +1,131 @@
+# CUI // SP-CTI
+# Phase 69: ProposalAI Module + Free Trials + Marketplace Admin
+
+## Overview
+
+Phase 69 adds three capabilities:
+- **Free Trial System** — 30-day trial for any marketplace module, renewable once (60 days max)
+- **ProposalAI Module** — Gates Phase 67/68 AI drafting features behind a marketplace subscription
+- **Marketplace Admin** — CRUD interface for managing modules and pricing tiers, plus redesigned marketplace page (1 card per module with pricing tiers inside)
+
+## Architecture Decisions
+
+| ID | Decision | Rationale |
+|----|----------|-----------|
+| D-P69-1 | New `trial` billing_cycle | Reuses existing subscription/license infrastructure |
+| D-P69-2 | Trial auto-activates, 30-day expiry | Same UX as free tier but time-limited |
+| D-P69-3 | `trial_renewal_count` column, max 1 renewal | 60 days max before requiring paid subscription |
+| D-P69-4 | Single `proposalai` module slug | One paywall for all AI drafting features |
+| D-P69-5 | `@require_module('proposalai')` on 8 API routes | Server-side 403 enforcement |
+| D-P69-6 | `{% if module_enabled('proposalai') %}` on templates | Client-side gating with upgrade prompts |
+| D-P69-7 | New `/marketplace/admin` page with CRUD | Admin configuration for modules + pricing |
+| D-P69-8 | Redesigned marketplace page — 1 card per module | Group pricing tiers inside card, not separate cards |
+
+## Schema Changes
+
+### ALTER: `marketplace_subscriptions`
+```sql
+ALTER TABLE marketplace_subscriptions ADD COLUMN trial_renewal_count INTEGER DEFAULT 0;
+```
+
+### CHECK Constraint Updates
+Both `marketplace_pricing` and `marketplace_subscriptions` tables updated to include `'trial'` in `billing_cycle` CHECK constraint.
+
+### New Seed Data
+- WriteGuard trial pricing tier ($0, 30 days)
+- ProposalAI asset (slug: `proposalai`, type: `skill`)
+- ProposalAI pricing: trial ($0), monthly ($149), annual ($1,490)
+
+## Modified Files
+
+| File | Changes |
+|------|---------|
+| `tools/db/init_icdev_db.py` | Trial ALTER SQL, CHECK constraint updates, ProposalAI + trial seed data |
+| `tools/marketplace/subscription_manager.py` | Trial in `create_subscription()`, new `renew_trial()` |
+| `tools/marketplace/license_manager.py` | Trial 30-day expiry in `issue_license()` |
+| `tools/dashboard/app.py` | 8 route gates, 2 trial routes, 5 admin CRUD routes, marketplace page redesign |
+| `tools/dashboard/templates/marketplace.html` | Redesign: 1 card per module with pricing tiers |
+| `tools/dashboard/templates/marketplace_admin.html` | New: Module + pricing CRUD admin |
+| `tools/dashboard/templates/marketplace_subscriptions.html` | Trial status, renewal button |
+| `tools/dashboard/templates/proposals/detail.html` | Gate Generate All/Generate buttons |
+| `tools/dashboard/templates/proposals/section_detail.html` | Gate Generate Draft, pipeline, KB panels |
+| `tools/dashboard/templates/govcon.html` | Gate KB nav link |
+
+## API Routes
+
+### Trial Routes
+| Route | Method | Purpose |
+|-------|--------|---------|
+| `/api/marketplace/trial/start` | POST | Start 30-day trial for a module |
+| `/api/marketplace/trial/renew` | POST | Renew trial once (another 30 days) |
+
+### Admin Routes
+| Route | Method | Purpose |
+|-------|--------|---------|
+| `/marketplace/admin` | GET | Admin page listing all modules + pricing |
+| `/api/marketplace/admin/assets` | POST | Create new module asset |
+| `/api/marketplace/admin/assets/<id>` | PUT | Update module |
+| `/api/marketplace/admin/pricing` | POST | Add pricing tier |
+| `/api/marketplace/admin/pricing/<id>` | PUT | Update pricing tier |
+| `/api/marketplace/admin/pricing/<id>` | DELETE | Soft deactivate pricing tier |
+
+### Gated Routes (ProposalAI)
+| Route | Purpose |
+|-------|---------|
+| `GET /govcon/knowledge-base` | KB browse page |
+| `POST /api/govcon/knowledge-base/upload` | KB document upload |
+| `GET /api/govcon/knowledge-base/search` | KB search |
+| `GET /api/govcon/knowledge-base/stats` | KB statistics |
+| `POST /api/govcon/sections/<id>/generate-draft` | Single section v2 pipeline |
+| `POST /api/govcon/sections/<id>/draft-status` | Draft status check |
+| `POST /api/govcon/proposals/<id>/generate-all` | Batch generation |
+| `GET /api/govcon/jobs/<id>/progress` | Batch job progress |
+
+## Trial Lifecycle
+
+```
+No subscription → "Start 30-Day Trial" button
+    ↓ POST /api/marketplace/trial/start
+Trial Active (30 days) → "Renew 30 Days" button
+    ↓ POST /api/marketplace/trial/renew
+Trial Active (extended 30 days) → "(renewal used)"
+    ↓ Expires or Cancel
+No subscription → "Start 30-Day Trial" / "Subscribe" buttons
+```
+
+## Marketplace Redesign
+
+**Before (Phase 66):** 1 card per pricing tier → 3 cards for WriteGuard (confusing)
+**After (Phase 69):** 1 card per module with pricing tiers listed inside
+
+Card layout:
+```
+┌────────────────────────────────────────┐
+│ ProposalAI — AI Proposal Drafting      │  From $149/mo
+│ proposalai · v1.0.0                    │
+│ Description text...                    │
+│ [govcon] [proposal] [drafting] [ai]    │
+│ ┌──────┐ ┌─────────┐ ┌──────────────┐ │
+│ │Trial │ │ Monthly │ │   Annual     │ │
+│ │ $0   │ │  $149   │ │  $1,490      │ │
+│ └──────┘ └─────────┘ └──────────────┘ │
+│ [Start Trial] [Subscribe] [Gov PO]     │
+│  — or — ✓ Trial Active (2026-04-03)   │
+└────────────────────────────────────────┘
+```
+
+## Verification Results
+
+| # | Check | Result |
+|---|-------|--------|
+| V1 | ProposalAI + trial tiers in DB | 2 assets, 7 pricing tiers |
+| V2 | Trial start | Trial activates, 30-day expiry, routes unlock (200) |
+| V3 | Trial renewal | Renew extends 30d, count=1, second attempt shows "(renewal used)" |
+| V4 | Route gating (no license) | 403 on KB/draft APIs |
+| V5 | Template gating | KB link hidden, Generate buttons hidden |
+| V6 | Cancel trial | Routes re-gated (403), buttons reappear on marketplace |
+| V7 | Marketplace redesign | 1 card per module, tiers inside, 3 viewports |
+| V8 | Admin CRUD | Modules + pricing displayed with Edit/Deactivate/Publish |
+| V9 | Subscriptions page | Trial status, renewal info, Cancel button |
+| V10 | Console errors | 0 errors across all pages |
+| V11 | Network 500s | 0 server errors |