Founder AI: Agentic OS & Observability

Multi-agent studio OS powered by Gemini & Google Cloud with real-time Grafana observability telemetry monitoring agent execution, policy guardrails, and USDC budget spending.

Comment

💡 Inspiration

Independent film producers, media studio founders, and executive directors face massive operational friction, crew/VFX vendor payment bottlenecks, and budget leakages across multi-stage creative workflows. In film production, managing spending guardrails while maintaining real-time visibility into complex agent reasoning pipelines requires a Studio Control Room—a system that not only executes multi-step strategic reasoning but provides complete observability and governance over every operational and financial decision.

Founder AI was built to empower media studio founders and executive producers to turn complex production challenges into structured execution using Google Cloud Gemini, backed by real-time Grafana Labs observability telemetry and cryptographically signed policy engines.

🎬 What it does

Founder AI is a closed-loop multi-agent executive operating system and studio command center tailored for media & entertainment operations:

  1. Multi-Agent Reasoning Pipeline: Orchestrates 7 specialized agents (AssessmentAgent, FrameworkSelectionAgent, KnowledgeRetrievalAgent, StrategyAgent, ExecutionCoachAgent, MemoryAgent, ResponseComposer) powered by Google Cloud Vertex AI & Gemini to diagnose production bottlenecks, budget overruns, and distribution agreement risks.
  2. Framework-Grounded Execution: Auto-selects from 13 proprietary operational frameworks (such as ECG KISS, SLR CAMERAS, RUN DCMS ER) to generate deterministic 8-part strategy reports for studio leadership.
  3. Grafana Studio Control Room (Grafana Labs Track Integration): Exports real-time Prometheus telemetry to Grafana to monitor agent step latencies, policy pass/fail rates, framework selection frequencies, and USDC production budget payouts.
  4. Deterministic Policy & Agentic Payouts: Executes automated USDC crew and vendor payouts via Circle / Razorpay backed by an HMAC-SHA256 signed PolicyEngine that blocks unbudgeted spending or prompt injection exploits during production transfers.

🛠️ How we built it

  • AI Core & Cloud: Google Cloud Vertex AI / Gemini 1.5 Pro for strategic reasoning and context composition; ChromaDB for local vector embeddings.
  • Grafana Labs Telemetry Track: Integrated prometheus_client across orchestrator.py, policy_engine.py, and payment_agent.py on port 9090 with pre-built Grafana studio dashboards (config/grafana_dashboard.json).
  • Client Desktop UI: Native PyQt6 cross-platform desktop application with live wallet balance widgets and real-time step status callbacks.
  • Governance & Security: HMAC-SHA256 signature verification in SQLite preventing manual database tampering from unlocking frozen production caps.

🚧 Challenges we faced

  • Studio Telemetry & Step Observability: Multi-agent chains in media workflows can feel like "black boxes." We solved this by integrating Prometheus metrics to track step-by-step latency histograms for every agent phase in Grafana.
  • Strict Response Contracts: Ensuring LLM outputs strictly follow an 8-part markdown contract without leaking prompt delimiters required custom deterministic verification and retry logic.
  • Financial Safety in Media Spend: Decoupling payment validation from the LLM core using a cryptographically signed Policy Engine to prevent hallucinations from triggering unauthorized funds transfers.

🏆 Accomplishments that we're proud of

  • 100% Observable Studio Control Room: Built a production-grade multi-agent engine with live Grafana dashboards tracking metrics down to sub-second agent execution step durations.
  • Zero-Trust Security Boundary: Cryptographically sealed spending policies with emergency kill-switch protection for studio budgets.
  • Complete Framework Alignment: Grounded reasoning strictly in 13 structured operational frameworks.

🎓 What we learned

  • How to export granular multi-agent execution telemetry to Prometheus and map it visually inside Grafana Cloud/Local dashboards.
  • Best practices for combining Google Cloud Gemini reasoning with deterministic local governance boundaries in media production environments.

🔮 What's next for Founder AI

  • Cloud Run Deployment: Exposing studio telemetry and agent endpoints via serverless Google Cloud Run microservices.
  • Multi-Tenant Grafana Cloud Integration: Streaming agent observability metrics to Grafana Cloud instances automatically for enterprise studio teams.

Built With

Share this project:

Updates

Submission history