Commit graph

8 commits

Author SHA1 Message Date
Vadym Samoilenko
f73291285d Improve presentation pipeline: brief summarization + section attribution + narrative continuity
Based on PPTAgent (EMNLP 2025) and DocPres research findings:

1. Brief summarization (summarize_brief.py)
   - For content >800 chars: single LLM call extracts {overview, sections[{title,
     key_points, data_points}]} before outline generation
   - Prevents "lost middle" context loss in long documents
   - BriefStructure.to_outline_context() formats sections for outline prompt
   - BriefStructure.get_section_text(idx) returns targeted excerpt per slide

2. Section attribution in SlideOutlineModel
   - Added source_section_idx: Optional[int] field
   - LLM sets this during outline generation to map each slide → brief section
   - Used to pass targeted section text to per-slide content generation
     instead of full brief (reduces hallucination, improves accuracy)

3. Narrative continuity in slide content generation
   - prev_slide_title passed to each content generation call
   - Injected in user prompt: "ensure this slide continues naturally from..."
   - Batch-safe: titles collected from completed batch before next starts

4. Source section text in content generation
   - source_section_text parameter added to get_slide_content_from_type_and_outline
   - Injected as "Source Material for This Slide" in user prompt
   - Only data points present in the excerpt should be used

5. Richer layout catalog
   - PresentationLayoutModel.to_catalog_string() added
   - Includes field names + maxLength constraints alongside layout descriptions
   - Helps LLM make informed layout choices based on content type

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-19 20:22:22 +00:00
Vadym Samoilenko
587f5ef6e1 Add 3 sandbox features: diagrams, mermaid, and template code-gen
Feature 1 — PPTX from Template (Code-Gen):
- backend/services/template_codegen_service.py: analyze PPTX, strip slides,
  Gemini code-gen + subprocess exec (60s timeout, auto-retry on error)
- backend/api/v1/ppt/endpoints/template_codegen.py: POST /template-codegen/generate
  (multipart: presentation_id + template_file + custom_prompt, rate-limited 3/min)
- frontend/components/TemplateCodegenExport.tsx: drag-drop modal
- Header.tsx: "Export from Template" option in export dropdown

Feature 2 — Diagrams in Slides:
- backend/models/diagram_data.py: DiagramData / FlowStep / BarChartItem models
- generate_slide_content.py: optional __diagram__ + __mermaid__ fields in LLM schema
- DiagramRenderer.tsx: pure React flowchart / bar chart / pie chart (no deps)
- SlideRenderer.tsx: chart elements render DiagramRenderer/MermaidRenderer;
  floating overlay fallback when no chart element exists in JSON layout
- V1ContentRender.tsx: diagram/mermaid overlay on built-in template slides
- generate-pptx/route.ts: addDiagramToSlide() — bar/pie via pptxgenjs addChart(),
  flowchart via addShape()+addText(), mermaid via /api/mermaid-to-image

Feature 3 — Mermaid Diagrams:
- MermaidRenderer.tsx: dynamic import mermaid@11, useEffect render, error fallback
- frontend/app/api/mermaid-to-image/route.ts: Puppeteer renders Mermaid to PNG → base64

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-19 18:47:31 +00:00
Vadym Samoilenko
ae41562103 Phase 8: Data-driven slide architecture + template management overhaul
Replaces TSX/Babel compilation pipeline with a JSON element model:
- New _do_parse_v2(): 1 LLM call/layout (vs 2) classifies OXML geometry
  elements into placeholder types → JSON stored in layout_code
- SlideRenderer.tsx: renders JSON element model as %-positioned divs,
  no Babel compilation or runtime errors
- parseLayoutSchema.ts: isJsonLayoutCode() / parseLayoutSchema() /
  mergeElementsWithContent() — full JSON schema parsing layer
- useCustomTemplates.ts: transparent dual-format support (JSON + TSX)
  via parsedLayoutToCompiled() adapter

Template management improvements:
- PresentationLayoutCodeModel: +is_enabled (bool) +thumbnail_path (str)
- Migration 005: adds both columns to presentation_layout_codes
- DELETE /master-decks/{id}: hard delete (files + TemplateModel +
  PresentationLayoutCodeModel rows + MasterDeckModel)
- PATCH /template-management/layouts/{db_id}/toggle-enabled: new endpoint
- LayoutData response: +db_id, +is_enabled, +thumbnail_path
- _register_as_template(): stores thumbnail_path + is_enabled per layout

Admin UI:
- /admin/templates/ — list all custom templates with delete
- /admin/templates/[id]/ — layout grid with screenshots + enable/disable
- AdminSidebar: Templates nav item

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-01 20:05:25 +00:00
Vadym Samoilenko
c431d4ab45 Implement critical security fixes and modern design system (Pre-launch P0 tasks)
Security Improvements (P0.0-P0.4):
- P0.0: Migrate to Gemini-only AI stack (simplified, single billing)
- P0.1: Fix CORS to restrict allowed origins from env (was *)
- P0.2: Remove hardcoded dev password, require env var
- P0.3: Add rate limiting (slowapi) - 3-10 req/min on sensitive endpoints
- P0.4: Add request size limits (100MB default via middleware)

New Features:
- Unified LLM service with Google Gemini priority
- OXML geometry extractor for layout parsing
- TSX validator for generated React components
- Client ID support in presentation requests with access control
- Configurable LLM/image timeouts via env vars

Modern Design System (P0.9 - partial):
- Enhanced CSS design tokens (primary, semantic colors, shadows)
- Typography scale (h1-h4, body variants, caption)
- Modern animations (fadeIn, slideIn, scaleIn)
- Updated Button component with better variants and hover effects
- Created unified Card and StatusBadge components
- Applied design system to Dashboard and Settings pages

Backend Improvements:
- Master deck parser simplification
- Slide-to-HTML endpoint cleanup (325 lines removed)
- Better error handling in prompts endpoint

Frontend Improvements:
- Settings UI simplified to show only Google/Gemini
- Dashboard uses CSS variables instead of hardcoded colors
- Improved button transitions and hover states

Co-Authored-By: Claude Sonnet 4.5 (1M context) <noreply@anthropic.com>
2026-02-27 18:28:24 +00:00
Vadym Samoilenko
69a8829750 Phase 3: Bug fixes, feature enhancements, and polish
P0 Critical: presentation isolation (client scoping), storage super_admin fix,
template selection in worker, IMAGE_PROVIDERS list fix.

P1 High: template layout management UI (delete/filter/bulk), slide-based parsing
mode, LLM model listing & connection test, settings persistence to DB (Fernet
encryption), logout button.

P2 Polish: storage improvements (master deck files, per-client breakdown, bulk
delete, hard purge, client selector), image generation error visibility
(__image_error__ badge), hamster wheel loading animation.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-27 12:58:52 +00:00
Vadym Samoilenko
d3d1667a79 Phase 2: Admin panel, analytics, storage, template pipeline, multi-provider LLM
- Fix admin sidebar: remove duplicate Teams, add Storage nav item
- Analytics: client-scoped queries, super_admin sees all (including NULL client_id)
- Storage management: list/download/delete presentations with file metadata
- Settings page with brand config router
- AI usage tracking: new AIUsageModel, ai_usage_service, analytics endpoint
- Master deck → template bridge: _register_as_template creates TemplateModel
  + PresentationLayoutCodeModel so parsed layouts appear in template picker
- Multi-provider LLM vision in parser: Anthropic/Google/OpenAI with asyncio.to_thread
- Fix PPTX upload 400: accept by .pptx extension (browser sends octet-stream)
- Fix reparse FK violation: presentation_id=None for parse_master_deck jobs
- Worker job_timeout increased to 1800s for LLM-heavy master deck parsing
- PYTHONUNBUFFERED=1 in docker-compose worker for real-time log output
- Auth: clientId in /me response, dev-login cookie improvements
- Frontend: auth slice clientId, master-deck thumbnails, storage page

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-26 23:39:34 +00:00
Vadym Samoilenko
a2bd4cfefa Phase 3: Content Pipeline — file parsing, content intelligence, slide mapping, native charts
- Step 10: Extended file upload for Excel/CSV/images/URLs (openpyxl, trafilatura)
- Step 11: Content intelligence service with rule-based + LLM classification
- Step 12: Slide mapping engine mapping content blocks to master deck layouts
- Step 13: Chart data extractor, native PPTX chart service (bar/line/pie/gantt/waterfall), ChartDataEditor skeleton

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-26 15:54:04 +00:00
Vadym Samoilenko
cf21ba4516 Phase 1-2: Foundation + Admin Panel & Client Management
Phase 1 (Foundation):
- Project restructure (presenton-main → backend/ + frontend/)
- Database schema (8 new models, Alembic config, seed script)
- Auth (Azure AD SSO + dev bypass, JWT sessions, AuthMiddleware)
- RBAC (access_service, rbac_middleware, admin routers)
- Audit logging (fire-and-forget, AuditMiddleware, admin router)
- i18n (react-i18next with 5 namespace files)

Phase 2 (Admin Panel & Client Management):
- Admin panel shell (sidebar layout, role guard, 12 pages)
- Redux admin slice with 18 async thunks
- User management (role changes, deactivation)
- Client management (CRUD, brand config, team management)
- Brand config editor (colors, fonts, logos, voice rules)
- Master deck upload & parser (PPTX → HTML → React pipeline)
- Audit log viewer with filters and CSV/JSON export

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-26 15:37:17 +00:00