How prarabdha7/gcici26 works
JA Assure is a multi-agent marketing and compliance system for three insurance brands. Its primary workflow accepts a campaign request, orchestrates research, drafting and compliance checks, generates media, and persists content for human review; approved content is then published by a scheduled worker. The README also describes B2B lead generation, competitive intelligence, memory, and an Obsidian knowledge vault. Relationships below rely on the supplied excerpts, SOURCE INDEX references, and README; unsampled implementation details are not inferred.
Main entry points: nodes.py Workflow Nodes · actions.py Action API · database.py SQL Database · client.py LLM Client
Components 28
API and Review
- Action API
app/api/actions.py - REST Views
app/api/routes.py - Review Dashboard
app/api/dashboard.py
Marketing Workflow
- Marketing Graph
app/graph/graph.py - Workflow Nodes
app/graph/nodes.py - Run State
app/graph/state.py - Compliance Fallback
app/llm/fallback.py - Brand Knowledge
app/agents/brand_knowledge.py
Research and Leads
- Lead Generation
app/graph/lead_gen_graph.py - Lead Agents
app/agents/leads.py - Market Research
app/agents/research.py - Web Research
app/utils/research.py - Search Providers
app/agents/providers.py - Intel Worker
worker/intel.py
Models and Media
- LLM Client
app/llm/client.py - Media Assembly
app/media/assembly.py - Image Generation
app/media/image_gen.py - Stock Video
app/media/stock_video.py
Memory and Publishing
- SQL Database
app/db/database.py - Content and Lead Records
app/db/models.py - Feedback Memory
app/memory/store.py - Memory Retrieval
app/memory/retrieval.py - Obsidian Vault
app/integrations/obsidian.py - Publish Scheduler
worker/scheduler.py - Social Publisher
worker/publisher.py
Other components
- Marketing User
- Model Providers
- Social Platforms
Connections 39
- Marketing User → Action API: submits requests
app/api/actions.py - Action API → Marketing Graph: runs graph
app/api/actions.py - Marketing Graph → Workflow Nodes: dispatches steps
app/graph/graph.py - Marketing Graph → Run State: uses state
app/graph/graph.py - Workflow Nodes → LLM Client: generates copy
app/graph/nodes.py - Workflow Nodes → Compliance Fallback: uses fallback
app/graph/nodes.py - Workflow Nodes → Memory Retrieval: retrieves guidance
app/graph/nodes.py - Workflow Nodes → Media Assembly: assembles media
app/graph/nodes.py - Workflow Nodes → Image Generation: generates images
app/graph/nodes.py - Workflow Nodes → SQL Database: persists queue
app/graph/nodes.py - Workflow Nodes → Brand Knowledge: uses brand rules
- Action API → Image Generation: serves image actions
app/api/actions.py - Action API → Media Assembly: serves media actions
app/api/actions.py - Action API → Obsidian Vault: uses vault
app/api/actions.py - Action API → SQL Database: reads and writes
app/api/actions.py - Action API → Feedback Memory: records feedback
app/api/actions.py - REST Views → SQL Database: queries records
app/api/routes.py - Review Dashboard → Marketing Graph: uses graph
app/api/dashboard.py - Review Dashboard → SQL Database: reads records
app/api/dashboard.py - Lead Generation → Lead Agents: dispatches discovery
app/graph/lead_gen_graph.py - Lead Generation → Market Research: dispatches research
app/graph/lead_gen_graph.py - Market Research → Web Research: fetches context
app/agents/research.py - Market Research → LLM Client: summarizes findings
app/agents/research.py - Lead Agents → Search Providers: discovers prospects
app/agents/leads.py - Lead Agents → SQL Database: stores leads
app/agents/leads.py - Web Research → Search Providers: fallback search
- Intel Worker → Market Research: runs research
worker/intel.py - LLM Client → Model Providers: calls models
app/llm/client.py - Media Assembly → LLM Client: requests generation
app/media/assembly.py - Media Assembly → Stock Video: selects footage
app/media/assembly.py - Image Generation → LLM Client: requests synthesis
app/media/image_gen.py - SQL Database → Content and Lead Records: maps tables
app/db/database.py - Memory Retrieval → Content and Lead Records: reads feedback
app/memory/retrieval.py - Feedback Memory → Content and Lead Records: stores feedback
app/memory/store.py - Publish Scheduler → Social Publisher: dispatches approved
worker/scheduler.py - Publish Scheduler → Intel Worker: schedules sweeps
worker/scheduler.py - Publish Scheduler → SQL Database: polls queue
worker/scheduler.py - Social Publisher → Social Platforms: publishes posts
worker/publisher.py - Social Platforms → REST Views: confirms webhook
app/api/routes.py