🔥 Prometheus: Enterprise Workstream Observability & Asynchronous Orchestration Platform

Submission for All Things Agentic Hackathon: The Fortified Enterprise Fleet Track
Autonomous AI Chief of Staff correlating cross-functional engineering telemetry (GitHub, Jira, Slack, CI/CD), isolating data via ABAC/RLS, and proposing contextual remediation actions for human sign-off ("Propose, Don't Impose") on the Gemini Enterprise Agent Platform.

🏛️ System Architecture & Sub-Agent Fleet

                          [ User Query / 08:00 AM Cron / Webhook / MCP ]
                                                │
                                                ▼
                                ┌───────────────────────────────────┐
                                │    Router & Guardrail Agent       │
                                │   - ABAC & Scope Perimeter Check  │
                                │   - Prompt Defense / PII Filter   │
                                └────────────────┬──────────────────┘
                                                 │
                         ┌───────────────────────┼──────────────────────┐
                         ▼                       ▼                      ▼
             ┌──────────────────────┐ ┌───────────────────┐ ┌──────────────────────┐
             │ Git & CI/CD Ingestion│ │  Project Tracker  │ │ Workstream Ingestion │
             │ - GitHub / GitLab    │ │  - Jira / Linear  │ │ - Public Slack/Teams │
             │ - PR latency / CI    │ │  - Epics/Blockers │ │ - Calendar Metadata  │
             └──────────┬───────────┘ └─────────┬─────────┘ └──────────┬───────────┘
                         │                      │                      │
                         └──────────────────────┼──────────────────────┘
                                                │ (Aggregated Telemetry via MCP)
                                                ▼
                                ┌───────────────────────────────────┐
                                │    Synthesis & Blocker Agent      │
                                │    - Gemini 3.7 Flash on Vertex AI│
                                │    - Enterprise Async Execution   │
                                │    - Root-Cause Bottleneck Finder │
                                └────────────────┬──────────────────┘
                                                 │
                                                 ▼
                                ┌───────────────────────────────────┐
                                │     Action & Drafting Agent       │
                                │     - "Propose, Don't Impose"     │
                                │     - Slack Block Kit Action Cards│
                                └────────────────┬──────────────────┘
                                                 │
                                                 ▼
                                   [ Human Approval Checkpoint ]
                                   (Approve / Edit / Discard)
                                                 │
                                     ┌───────────┴───────────┐
                                     ▼                       ▼
                            [ Approved: Dispatch ]   [ Rejected: Abort ]
                             - Slack DM / Jira SDK

🌟 Enterprise Fleet Architectural Pillars

Pillar	Platform Implementation	Capabilities
Discovery & Lifecycle	Agent Registry (`app/registry/`)	Central cataloging, schema validation, tool binding, and versioning for all 6 sub-agents.
Core Execution & State	Async DAG + Memory Bank (`app/memory/`)	Checkpointing, multi-week blocker lifecycles, SQLite persistence, and human approval state machines.
Security & Governance	ABAC Perimeter & Model Armor (`app/security/`)	Deterministic scope validation $P(U,R)$, PII redaction (tokens, emails, phone numbers), and prompt injection filters.
Observability & Protocols	Hybrid MCP + Structured Traces (`app/mcp/`)	Dual `stdio` & `SSE` Model Context Protocol transports + OpenTelemetry trace headers.

⚡ Vertex AI & Gemini Enterprise Agent Platform

Prometheus is unified exclusively on Gemini 3.7 Flash powered by Google Cloud Vertex AI & the Gemini Enterprise Agent Platform:

Enterprise IAM & ADC: Uses native Google Cloud Application Default Credentials (ADC) and IAM service accounts with zero raw API keys exposed.
Non-Blocking Async Engine: Native client.aio.models.generate_content() execution with timeout guards preventing event loop starvation.
Telemetry Delta Caching: Inputs are hashed (SHA-256) to return cached digests for unchanged telemetry, saving token consumption and latency.
Agent Engine Hosting: Designed to deploy natively onto the Gemini Enterprise Agent Platform (Agent Engine).

📁 Repository Layout

prometheus/
├── app/
│   ├── __init__.py
│   ├── config.py                 # Gemini 3.7 Flash, Vertex AI & Agent Engine settings
│   ├── main.py                   # FastAPI REST API, MCP SSE Stream & CLI runner
│   ├── scheduler.py              # Background 08:00 AM async cron scheduler
│   ├── agents/
│   │   ├── __init__.py
│   │   ├── router_agent.py       # ABAC & Guardrail security perimeter
│   │   ├── git_agent.py          # GitHub/CI ingestion agent
│   │   ├── jira_agent.py         # Jira/Linear project tracker agent
│   │   ├── workstream_agent.py   # Slack & calendar ingestion agent
│   │   ├── synthesis_agent.py    # Multi-domain correlation (Gemini 3.7 Flash)
│   │   └── action_agent.py       # HITL action drafter with Slack Block Kit
│   ├── llm/
│   │   ├── __init__.py
│   │   └── gemini_pool.py        # Vertex AI Gemini 3.7 Flash engine & cache
│   ├── mcp/
│   │   ├── __init__.py
│   │   ├── protocol.py           # MCP JSON-RPC 2.0 schemas
│   │   ├── server.py             # Prometheus MCP Server (stdio & SSE)
│   │   └── client.py             # MCP Client adapter for external tools
│   ├── registry/
│   │   ├── __init__.py
│   │   └── agent_registry.py     # Sub-Agent cataloging & discovery
│   ├── tools/
│   │   ├── __init__.py
│   │   ├── github_tools.py       # Git PR & CI status tools
│   │   ├── jira_tools.py         # Project management tools
│   │   └── slack_tools.py        # Communication drafting & dispatch tools
│   ├── security/
│   │   ├── __init__.py
│   │   ├── abac_guard.py         # Attribute-Based Access Control (ABAC/RLS)
│   │   └── guardrails.py         # Inline prompt defense & PII sanitizer
│   ├── memory/
│   │   ├── __init__.py
│   │   └── state_store.py        # Durable SQLite session checkpointing & audit store
│   └── workflows/
│       ├── __init__.py
│       └── prometheus_flow.py    # Multi-agent asynchronous orchestration DAG
├── tests/
│   ├── mock_telemetry.py         # Test fixtures & telemetry generators
│   └── test_workflow.py          # Hermetic multi-agent & security test suite
├── .github/
│   └── workflows/ci.yml          # GitHub Actions automated CI pipeline
├── deploy_gcp.sh                 # Deployment script for Gemini Enterprise Agent Engine
├── Dockerfile                    # Multi-stage production containerfile
├── Procfile                      # PaaS process definition
├── requirements.txt              # Production & test dependencies
├── PROJECT_SPEC.md               # Detailed system design specification
└── .env.example                  # Environment variables template

🚀 Quickstart & Setup

1. Installation

# Clone repository
git clone https://github.com/Tony-Stark2025/prometheus.git
cd prometheus

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

2. Configure Environment

cp .env.example .env
# Authenticate with Google Cloud
gcloud auth application-default login

3. Run the Interactive CLI Demo

python app/main.py --cli

4. Run the REST API & MCP Server

uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload

OpenAPI Interactive Documentation: http://localhost:8000/docs
MCP Server-Sent Events Endpoint: http://localhost:8000/mcp/sse
Agent Registry Discovery: http://localhost:8000/api/v1/registry/agents

🌐 Live Google Cloud Deployments

Component	Target Infrastructure	Live GCP Resource / URL	Status
Multi-Agent Reasoning Fleet	Vertex AI Agent Engine (`us-central1`)	`projects/135010851380/locations/us-central1/reasoningEngines/954065480874721280`	`ACTIVE / SERVING`
FastAPI Backend & MCP Server	Google Cloud Run (`us-central1`)	prometheus-chief-of-staff-135010851380.us-central1.run.app	`ACTIVE / 100% TRAFFIC`
Interactive Executive Dashboard	Cloud Run Web UI	prometheus-chief-of-staff-135010851380.us-central1.run.app/dashboard	`200 OK`
Model Context Protocol (MCP)	Cloud Run SSE Endpoint	`POST https://prometheus-chief-of-staff-135010851380.us-central1.run.app/mcp/sse`	`JSON-RPC 2.0`
Foundation Model	Gemini 3.7 Flash	Standardized Vertex AI Model	`SERVING`

🔌 Connecting to Claude Desktop / Antigravity via MCP

Add Prometheus to your MCP client configuration (e.g. claude_desktop_config.json):

{
  "mcpServers": {
    "prometheus-cloud": {
      "url": "https://prometheus-chief-of-staff-135010851380.us-central1.run.app/mcp/sse"
    },
    "prometheus-local": {
      "command": "python",
      "args": ["-m", "prometheus.mcp.server"]
    }
  }
}

☁️ Deployment on Google Cloud (Agent Engine & Cloud Run)

1. Deploy Reasoning Engine to Vertex AI Agent Engine:

python deploy_agent_engine.py --project gen-lang-client-0942141479 --location us-central1

2. Deploy Container to Google Cloud Run:

gcloud run deploy prometheus-chief-of-staff \
    --source . \
    --region=us-central1 \
    --allow-unauthenticated \
    --port=8080 \
    --set-env-vars="USE_VERTEX_AI=true,GCP_PROJECT_ID=gen-lang-client-0942141479,GCP_LOCATION=us-central1,AGENT_ENGINE_APP_ID=projects/135010851380/locations/us-central1/reasoningEngines/954065480874721280,ENVIRONMENT=production,MCP_ENABLED=true"

3. Verify Live Deployments Programmatically:

# Verify live Vertex AI Reasoning Engine
python deploy_agent_engine.py --verify-only projects/135010851380/locations/us-central1/reasoningEngines/954065480874721280

# Verify full 157-test matrix
pytest -v

🔐 Google Secret Manager Integration

Configure secrets securely via Google Secret Manager URIs in .env or Cloud Run environment variables:

GITHUB_TOKEN=sm://prometheus-github-token
JIRA_API_TOKEN=sm://prometheus-jira-token
SLACK_BOT_TOKEN=sm://prometheus-slack-token

Prometheus automatically resolves sm://<secret_id> directly from Google Cloud Secret Manager at startup with zero secret exposure in logs.

🧪 Testing

Run the comprehensive 5-tier test matrix:

pytest tests/ -v

All 157 tests execute hermetically with zero mock leakage.

Name		Name	Last commit message	Last commit date
.agents		.agents
.github/workflows		.github/workflows
app		app
docs		docs
prometheus		prometheus
tests		tests
tools		tools
.env.example		.env.example
.gcloudignore		.gcloudignore
.gitignore		.gitignore
DEVPOST_SUBMISSION.md		DEVPOST_SUBMISSION.md
Dockerfile		Dockerfile
LICENSE		LICENSE
ORIGINAL_REQUEST.md		ORIGINAL_REQUEST.md
PROJECT.md		PROJECT.md
PROJECT_SPEC.md		PROJECT_SPEC.md
Procfile		Procfile
README.md		README.md
TEST_INFRA.md		TEST_INFRA.md
deploy_agent_engine.py		deploy_agent_engine.py
deploy_gcp.ps1		deploy_gcp.ps1
deploy_gcp.sh		deploy_gcp.sh
prometheus_exam_portal.html		prometheus_exam_portal.html
pytest.ini		pytest.ini
refactor_namespace.py		refactor_namespace.py
requirements.txt		requirements.txt
setup.py		setup.py

Navigation Menu

Repository files navigation

🔥 Prometheus: Enterprise Workstream Observability & Asynchronous Orchestration Platform

🏛️ System Architecture & Sub-Agent Fleet

🌟 Enterprise Fleet Architectural Pillars

⚡ Vertex AI & Gemini Enterprise Agent Platform

📁 Repository Layout

🚀 Quickstart & Setup

1. Installation

2. Configure Environment

3. Run the Interactive CLI Demo

4. Run the REST API & MCP Server

🌐 Live Google Cloud Deployments

🔌 Connecting to Claude Desktop / Antigravity via MCP

☁️ Deployment on Google Cloud (Agent Engine & Cloud Run)

1. Deploy Reasoning Engine to Vertex AI Agent Engine:

2. Deploy Container to Google Cloud Run:

3. Verify Live Deployments Programmatically:

🔐 Google Secret Manager Integration

🧪 Testing

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

🔥 Prometheus: Enterprise Workstream Observability & Asynchronous Orchestration Platform

🏛️ System Architecture & Sub-Agent Fleet

🌟 Enterprise Fleet Architectural Pillars

⚡ Vertex AI & Gemini Enterprise Agent Platform

📁 Repository Layout

🚀 Quickstart & Setup

1. Installation

2. Configure Environment

3. Run the Interactive CLI Demo

4. Run the REST API & MCP Server

🌐 Live Google Cloud Deployments

🔌 Connecting to Claude Desktop / Antigravity via MCP

☁️ Deployment on Google Cloud (Agent Engine & Cloud Run)

1. Deploy Reasoning Engine to Vertex AI Agent Engine:

2. Deploy Container to Google Cloud Run:

3. Verify Live Deployments Programmatically:

🔐 Google Secret Manager Integration

🧪 Testing

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages