How nazazx/casevisionb works
Nomad Roasters Content Agent turns a plain-language brief and one server-selected promotion snapshot into a validated Instagram caption—or a blocked/clarification result. CLI and Streamlit are the user-facing entry points; the pipeline coordinates planning and finalization, while catalog policy keeps product and commercial facts authoritative. The README documents hybrid retrieval, optional brand knowledge, audit records, and clarification selection. Some source excerpts are partial, so uncertain internal wiring is omitted.
Main entry points: pipeline.py Content pipeline · catalog.py Snapshot catalog · cli.py CLI · llm.py Caption planner
Components 17
User interfaces
- CLI
src/nomad_agent/cli.py - Streamlit dashboard
app.py
Generation workflow
- Content pipeline
src/nomad_agent/pipeline.py - Caption planner
src/nomad_agent/llm.py - Plan normalization
src/nomad_agent/policy.py - Caption validation
src/nomad_agent/renderer.py - Approved phrase catalogue
src/nomad_agent/phrases.py
Product and brand facts
- Snapshot catalog
src/nomad_agent/catalog.py - Product matching
src/nomad_agent/product_retrieval.py - Promotion feed
promotions_feed_week1.json - Brand guidance
brand_kit.md - Brand knowledge search
src/nomad_agent/brand.py
Results and integrations
- Generation result
src/nomad_agent/schemas.py - Run audit log
src/nomad_agent/audit.py - Run telemetry
src/nomad_agent/telemetry.py
Other components
- Content creator
- OpenAI API
Connections 21
- Content creator → CLI: submits brief
- Content creator → Streamlit dashboard: submits brief
- CLI → Content pipeline: starts generation
- Streamlit dashboard → Content pipeline: starts generation
- CLI → Snapshot catalog: loads feed snapshot
- Streamlit dashboard → Snapshot catalog: selects feed snapshot
- Content pipeline → Caption planner: requests plan
- Caption planner → Snapshot catalog: searches products
- Caption planner → Snapshot catalog: loads product details
- Snapshot catalog → Product matching: ranks candidates
- Product matching → OpenAI API: embeds query and products
- Snapshot catalog → Promotion feed: reads snapshot
- Caption planner → Brand knowledge search: searches style references
- Brand knowledge search → OpenAI API: queries vector store
- Content pipeline → Plan normalization: normalizes plan
- Content pipeline → Caption validation: validates and renders
- Caption validation → Approved phrase catalogue: uses templates
- Content pipeline → Generation result: constructs result
- Content pipeline → Run telemetry: records run metrics
- CLI → Run audit log: appends run
- CLI → Brand guidance: loads guidance