REST API + Microsoft Teams Bot powered by a supervisor-style multi-agent LLM orchestrator.
Automates backlog engineering, architecture diagrams, BPMN processes, documents, presentations, web research, and data visualization — all from a single conversational interface.
╔══════════════════════════════════════════════════════════════════════════╗
║ ENTRY POINTS ║
║ ║
║ 👥 Microsoft Teams 🌐 External App / Swagger UI ║
║ (group chat or DM) (REST client, automation, testing) ║
║ │ │ ║
║ ▼ ▼ ║
║ Azure Bot Service X-API-Key header ║
║ (JWT validation) ║
║ │ │ ║
║ ▼ ▼ ║
║ POST /api/messages POST /api/v1/chat ║
╚══════════════════════════════════════════════════════════════════════════╝
│ │
└────────┬───────┘
▼
╔══════════════════════════════════════════════════════════════════════════╗
║ 🧠 CHIEF ARCHITECT — frysda ║
║ (LangGraph ReAct Supervisor) ║
║ ║
║ ┌─ Knowledge Base Search (BEFORE every response) ──────────────────┐ ║
║ │ Reads prior domain context, decisions, terminology │ ║
║ └───────────────────────────────────────────────────────────────────┘ ║
║ ║
║ EXECUTION MODES (auto-detected from intent): ║
║ ┌──────────────────────────────────────────────────────────────────┐ ║
║ │ MODE 1 │ User Story Engineering → JIRA + DOCS │ ║
║ │ MODE 2 │ Heuristic Audit → JIRA + MIRO + SLIDES │ ║
║ │ MODE 3 │ Discovery & Architecture→ WEB + PROCESS + ARCHITECT │ ║
║ │ MODE 4 │ Mass Change Management → JIRA + DOCS + PROCESS │ ║
║ │ MODE 5 │ Ad-hoc Orchestration → Intelligent routing │ ║
║ │ MODE 6 │ Data Visualization → CHARTS + WEB/JIRA │ ║
║ └──────────────────────────────────────────────────────────────────┘ ║
╚══════════════════════════════════════════════════════════════════════════╝
│ │ │ │ │ │ │ │
▼ ▼ ▼ ▼ ▼ ▼ ▼ ▼
┌────────┐ ┌────────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌────┐ ┌────┐ ┌──────┐
│ JIRA │ │ DOCS │ │SLIDE │ │ARCHI │ │PROCE │ │WEB │ │MIRO│ │CHART │
│ Agent │ │ Agent │ │ Agent│ │ TECT │ │ SS │ │Agt │ │Agt │ │ Agent│
└───┬────┘ └───┬────┘ └──┬───┘ └──┬───┘ └──┬───┘ └──┬─┘ └──┬─┘ └──┬───┘
│ │ │ │ │ │ │ │
▼ ▼ ▼ ▼ ▼ ▼ ▼ ▼
Atlassian Word/PDF PowerPoint Draw.io Camunda Tavily Miro PNG
Jira API .docx .pptx .drawio BPMN Search Board Chart
│
▼
╔══════════════════════════════════════════════════════════════════════════╗
║ 📚 KNOWLEDGE BASE (Auto-Learn) ║
║ After EVERY interaction: extracts & persists domain decisions, ║
║ user preferences, project terminology → business_output/knowledge_base ║
╚══════════════════════════════════════════════════════════════════════════╝
| # | Agent | Capabilities | Output |
|---|---|---|---|
| 1 | 🎫 JIRA | Create/update issues, User Stories, sprints, backlog engineering | Atlassian Jira |
| 2 | 📄 DOCS | Word documents, PDF, structured reports | .docx / .pdf |
| 3 | 📊 SLIDES | PowerPoint presentations, executive decks | .pptx |
| 4 | 🏗️ ARCHITECT | Architecture diagrams (hexagonal, C4, layered, cloud) | .drawio |
| 5 | 🔄 PROCESS | BPMN process models, swimlanes, Camunda deployment | .bpmn |
| 6 | 🌐 WEB | Web research, URL extraction, Tavily search | Markdown text |
| 7 | 🧩 MIRO | Brainstorming boards, sticky notes, mind maps | Miro board |
| 8 | 📈 CHARTS | Bar, line, pie, scatter charts, data visualization | .png |
Developer Machine
─────────────────────────────────────────────────────────────────────────
Terminal 1: API Server Terminal 2: Dev UI (optional)
┌─────────────────────────────┐ ┌─────────────────────────────┐
│ python api_server.py │ │ python dev_ui/server.py │
│ │ │ │
│ ✅ Startup events: │ │ ✅ Mock Teams-like UI │
│ · load .env │ │ · Simulates group chat │
│ · init GPT-4o via OpenAI │ │ · @frysda mention support │
│ · create orchestrator │ │ · Knowledge base viewer │
│ · register 8 agents │ └──────────────┬──────────────┘
│ · init Bot adapter (Teams) │ │ HTTP
│ │ ▼
│ 📡 Listening on :8000 │ http://localhost:8080
└────────────┬────────────────┘
│
┌─────────▼──────────────────────────┐
│ http://localhost:8000/docs │ ← Swagger UI
│ http://localhost:8000/health │ ← Health probe
│ http://localhost:8000/api/v1/chat │ ← Orchestrator entry
└─────────┬──────────────────────────┘
│
┌─────────▼───────────────────────────────────────────────────── ─┐
│ POST /api/v1/chat │
│ │
│ Request body: │
│ { │
│ "conversation_id": "session-01", │
│ "message": "Crie um diagrama de arquitetura hexagonal", │
│ "user_name": "Nilton" │
│ } │
│ │
│ Headers: X-API-Key: <your-api-key> (optional locally) │
└─────────┬───────────────────────────────────────────────────── ─┘
│
▼
┌───────────────────────────────────────────────────────────────── ┐
│ 🧠 ChiefArchitect.invoke() │
│ │
│ 1. search_knowledge_base("arquitetura hexagonal") │
│ → retrieves prior project context (if any) │
│ │
│ 2. Analyzes intent → MODE 3 (Discovery & Architecture) │
│ Presents execution plan to user │
│ │
│ 3. Delegates: │
│ · delegate_to_web → researches hexagonal architecture │
│ · delegate_to_architect → generates .drawio file │
│ │
│ 4. Composes final response │
│ │
│ 5. _auto_learn() → extracts & saves domain knowledge │
│ → business_output/knowledge_base/<hash>.txt │
└─────────┬────────────────────────────────────────────────────── ┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ Response JSON: │
│ { │
│ "response": "## Diagrama Gerado\n...", │
│ "conversation_id": "session-01", │
│ "files": ["diagrams/order_management_hexagonal.drawio"] │
│ } │
└─────────────────────────────────────────────────────────────────┘
Microsoft Teams
─────────────────────────────────────────────────────────────────────────
👤 User types in Teams group:
"@frysda crie uma User Story para o módulo de pagamentos"
│
▼
┌────────────────────────────────────────────────────────────────────┐
│ Microsoft Teams Client │
│ Detects @frysda mention → routes to registered bot │
└────────────────────────┬───────────────────────────────────────────┘
│ Teams internal routing
▼
┌────────────────────────────────────────────────────────────────────┐
│ Azure Bot Service │
│ │
│ · Validates bot registration (TEAMS_APP_ID) │
│ · Signs request with JWT (RS256) │
│ · Delivers Activity payload to Messaging Endpoint: │
│ POST https://<app>.azurecontainerapps.io/api/messages │
└────────────────────────┬───────────────────────────────────────────┘
│ HTTPS POST (JWT signed)
▼
┌────────────────────────────────────────────────────────────────────┐
│ Azure Container App — api_server.py │
│ Port 8000 │
│ │
│ POST /api/messages │
│ ↓ │
│ BotFrameworkAdapter.process_activity() │
│ · Validates JWT signature against Microsoft JWKS │
│ · Deserializes Teams Activity object │
│ ↓ │
│ BusinessBot.on_message_activity() │
│ · Extracts: conversation_id, user_text, user_name │
│ · Sends typing indicator back to Teams │
│ · Loads conversation history (CosmosDB or in-memory) │
│ ↓ │
│ _orchestrator.invoke(user_text, chat_history=history) │
└────────────────────────┬───────────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────────────────┐
│ 🧠 ChiefArchitect — same as local flow │
│ │
│ 1. search_knowledge_base(user_text) │
│ 2. Intent classification → MODE 1 (User Story Engineering) │
│ 3. Presents execution plan │
│ 4. delegate_to_jira: │
│ · Principal BA agent drafts complete User Story │
│ · Validates acceptance criteria, BDD scenarios │
│ · Creates issue in Atlassian Jira via API │
│ 5. _auto_learn() → persists domain knowledge │
│ (Azure File Share: /app/business_output/knowledge_base/) │
└────────────────────────┬───────────────────────────────────────────┘
│
▼
┌────────────────────────────────────────────────────────────────────┐
│ BusinessBot.on_message_activity() │
│ │
│ · Appends HumanMessage + AIMessage to history │
│ · Saves updated history to CosmosDB (or in-memory) │
│ · turn_context.send_activity(MessageFactory.text(response)) │
└────────────────────────┬───────────────────────────────────────────┘
│ Bot Framework reply
▼
┌────────────────────────────────────────────────────────────────────┐
│ Azure Bot Service │
│ Routes response back to originating Teams conversation │
└────────────────────────┬───────────────────────────────────────────┘
│
▼
👥 All group members see frysda's response in Teams chat
with formatted Markdown, tables, and action items
| Method | Endpoint | Auth | Description |
|---|---|---|---|
GET |
/health |
— | Azure liveness / readiness probe |
POST |
/api/messages |
Bot Framework JWT | Microsoft Teams webhook — only used by Azure Bot Service |
POST |
/api/v1/chat |
X-API-Key |
Send a message to the orchestrator |
DELETE |
/api/v1/conversations/{id} |
X-API-Key |
Clear conversation history for a session |
POST |
/api/v1/ingest |
X-API-Key |
Ingest a URL into the knowledge base |
GET |
/api/v1/files/charts/{filename} |
X-API-Key |
Download a generated PNG chart |
GET |
/api/v1/files/bpmn/{filename} |
X-API-Key |
Download a generated BPMN file |
GET |
/api/v1/files/diagrams/{filename} |
X-API-Key |
Download a generated Draw.io diagram |
Interactive docs:
/docs(Swagger UI) ·/redoc(ReDoc)
The orchestrator automatically learns from every interaction:
User message + Agent response
│
▼
_auto_learn()
│
GPT-4o extracts:
· Domain decisions made
· User preferences
· Project terminology
· Recurring patterns
│
▼
add_knowledge_entry()
│
business_output/knowledge_base/<hash>.txt
│
▼
Available on next startup via search_knowledge_base()
Local: stored at
./business_output/knowledge_base/
Azure: persisted on Azure File Share mounted at/app/business_output/knowledge_base/
# 1. Clone & install
git clone https://github.com/filhormnilton/BusinessAgent
cd BusinessAgent
pip install -r requirements.txt
# 2. Configure
cp .env.example .env
# Edit .env — minimum required: OPENAI_API_KEY
# 3a. Start API server + Swagger
python api_server.py
# → http://localhost:8000/docs
# 3b. Start Dev UI (Teams simulator)
python dev_ui/server.py
# → http://localhost:8080docker compose up --build
# → http://localhost:8000cp .env.example .env| Variable | Required | Description |
|---|---|---|
OPENAI_API_KEY |
✅ | OpenAI API key (GPT-4o) |
TEAMS_APP_ID |
Teams only | Azure Bot App ID |
TEAMS_APP_PASSWORD |
Teams only | Azure Bot client secret |
API_KEY |
Recommended | REST API auth key (X-API-Key header) |
JIRA_SERVER |
Jira features | Jira instance URL |
JIRA_USER |
Jira features | Jira user email |
JIRA_API_TOKEN |
Jira features | Jira API token |
JIRA_PROJECT_KEY |
Jira features | Default project key(s), comma-separated |
MIRO_ACCESS_TOKEN |
Miro features | Miro OAuth token |
MIRO_BOARD_ID |
Miro features | Target Miro board ID |
TAVILY_API_KEY |
Web search | Tavily search API key |
CAMUNDA_REST_URL |
BPMN deploy | Camunda engine REST URL |
COSMOS_ENDPOINT |
Recommended | CosmosDB for persistent chat history |
COSMOS_KEY |
Recommended | CosmosDB primary key |
BUSINESS_OUTPUT_DIR |
Optional | Output directory (default: ./business_output) |
| Resource | Purpose |
|---|---|
| Azure Container Registry (ACR) | Docker image storage |
| Azure Container App | Runs the API server |
| Azure Bot Service | Teams channel integration |
| Azure File Share | Persistent business_output/ |
| Azure CosmosDB (optional) | Persistent conversation history |
The workflow at .github/workflows/deploy.yml is triggered manually until Azure secrets are configured.
Required GitHub Secrets:
AZURE_CREDENTIALS → Service principal JSON
REGISTRY_LOGIN_SERVER → e.g. businessagentregistry.azurecr.io
REGISTRY_USERNAME → ACR admin username
REGISTRY_PASSWORD → ACR admin password
AZURE_RESOURCE_GROUP → Resource group name
CONTAINER_APP_NAME → Container App name
Deploy:
GitHub → Actions → Build & Deploy to Azure Container Apps → Run workflow
After deploy, set in Azure Bot Resource → Settings → Messaging endpoint:
https://<your-app>.azurecontainerapps.io/api/messages
BusinessAgent/
├── api_server.py ← Production entrypoint (FastAPI + Teams webhook)
├── business_main.py ← Local CLI for testing (not used in Azure)
├── Dockerfile
├── docker-compose.yml
├── requirements.txt
├── .env.example
├── .github/
│ └── workflows/
│ └── deploy.yml ← CI/CD to Azure Container Apps
└── Business/
├── config.py ← Centralized configuration (env vars)
├── agents/
│ ├── base.py
│ ├── jira_agent.py
│ ├── docs_agent.py
│ ├── slides_agent.py
│ ├── architect_agent.py
│ ├── process_agent.py
│ ├── web_agent.py
│ ├── miro_agent.py
│ └── charts_agent.py
├── mcp/
│ ├── api_jira.py
│ ├── api_knowledge_base.py
│ ├── api_drawio.py
│ ├── api_powerpoint.py
│ ├── api_miro.py
│ ├── api_web.py
│ ├── api_charts.py
│ └── api_office_pdf.py
├── orchestrator/
│ └── chief_architect.py ← LangGraph supervisor (frysda persona)
└── teams_bot/
├── bot.py ← BusinessBot activity handler
└── history_store.py ← In-memory or CosmosDB history
frysda · Chief Business Architect · Powered by GPT-4o + LangGraph
- Push to
mainbranch on GitHub - Configure secrets in GitHub → Settings → Secrets and variables → Actions
- Run the workflow manually: Actions → Build & Deploy to Azure Container Apps → Run workflow
- Get the Container App URL from Azure Portal
- Set in Azure Bot → Configuration → Messaging endpoint:
https://<app-url>/api/messages - Connect to Microsoft Teams channel in Azure Bot → Channels