No description
7e82a535a9
Replaces the one-shot form with a Copygen-style chat at /conversations/:id. Each turn is classified as generate / refine / clarify by a lightweight LLM intent router; generate and refine intents enqueue an RQ job that produces a new BannerSet, while clarify returns an inline reply without touching banners. New backend: - Conversation + ConversationMessage models + migration 0005 - intent_router service (chat_structured, 3-intent schema) - chat_turn RQ task with _persist_banner_set helper extracted from _generate_copy_async for reuse - /api/conversations CRUD + POST /messages endpoint - JobType.CHAT_TURN added New frontend: - ChatBrief page: message bubbles, inline BannerPreview cards with checkbox selection and "Open banner editor" CTA (same Medium+Large validation rule as VariantsGrid) - ConversationLanding: /conversations/new creates and redirects - conversationId added to journey store - "New Brief" nav now points to /conversations/new - Default route redirects to /conversations/new Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> |
||
|---|---|---|
| apache | ||
| assets/illustrations | ||
| backend | ||
| deploy | ||
| frontend | ||
| rag-corpus | ||
| .env.example | ||
| .gitignore | ||
| CLAUDE.md | ||
| deploy.sh | ||
| docker-compose.prod.yml | ||
| docker-compose.yml | ||
| README.md | ||
Barclays BMB Banner Visualiser
AI-powered banner creation tool for Barclays Mobile Banking campaign managers.
Brief → compliant copy variants → DAM image + icon → in-situ preview → PDF contact sheet + Workfront CSV
Live: https://optical-dev.oliver.solutions/barclays-banner-builder/
Stack
| Layer | Tech |
|---|---|
| Frontend | React 18 + Vite + TypeScript + TailwindCSS + shadcn/ui |
| Backend | FastAPI (Python 3.12) + SQLAlchemy 2.0 async |
| Queue | RQ + Redis (HTTP polling — no WebSocket) |
| DB | PostgreSQL 16 + pgvector |
| AI | OpenAI gpt-5.4-mini + text-embedding-3-small (RAG) |
| WeasyPrint | |
| Deploy | Docker Compose + Apache reverse proxy |
Quick start (dev)
cp .env.example .env
# Set OPENAI_API_KEY in .env
docker compose up -d --build
docker compose exec api alembic upgrade head
docker compose exec api python scripts/seed_admin.py
docker compose exec api python scripts/ingest_rag.py
docker compose exec api python scripts/index_icons.py
- Frontend: http://localhost:5174
- API docs: http://localhost:8010/docs
- Login:
admin@barclays.com/change_me_password
Deploy (Ubuntu + Apache)
# Clone once:
git clone <repo> /opt/barclays-banner-builder
# Every deploy (idempotent):
bash /opt/barclays-banner-builder/deploy.sh
# Options:
bash deploy.sh --reindex # force re-embed RAG + icons
bash deploy.sh --skip-frontend # backend-only update
The script handles: Docker build (with cache), DB migrations, first-run seeding, npm build, Apache config patching.
Key rules
- No WebSocket — all async work uses
POST → 202 + job_id → GET /api/jobs/{id}polling - OpenAI only — model
gpt-5.4-mini, key in.envonly (never committed) - System prompt editable from
/admin— full version history kept for FCA audit - Character limits enforced server-side — short_title ≤ 32, long_body ≤ 128, cta ≤ 50
- DAM mock mode — if
ADOBE_DAM_API_BASE_URLis empty, mock client is used automatically
Project structure
backend/app/
api/ — FastAPI routers
models/ — SQLAlchemy models
services/ — copy_generation, exporter, adobe_dam_client
llm/ — OpenAI client wrapper
rag/ — pgvector cosine retrieval
workers/ — RQ tasks (generate_copy, render_pdf)
prompts/ — system_prompt.md (initial seed)
frontend/src/
pages/ — BriefEditor, VariantsGrid, BannerEditor, ExportPage, AdminPage, LoginPage
lib/ — jobPolling.ts
store/ — Zustand auth store
rag-corpus/ — Consumer Duty, Consumer Understanding, Regulators .docx
assets/ — illustrations/ (PNG icons) + templates/
deploy/ — apache-barclays.conf (Apache Include fragment)
scripts/ — index_icons.py, ingest_rag.py, seed_admin.py