FLUME

V3 Autonomous Edge Orchestrator

An AI-powered agent workflow platform for planning, implementing, testing, and reviewing codebase changes natively using local hardware clusters.

---

The Flume V3 ecosystem is managed entirely by a singular, high-performance Go CLI (flume). This executable provides a pristine, zero-dependency interface to your terminal while cleanly orchestrating a containerized matrix (Elasticsearch, OpenBao, Python Workers) protecting your absolute host OS entirely.

✨ V3.2 Feature Highlights

• 10+ Native CLI Mappings: Deeply configure, execute RAG tasks, monitor statuses, and pull container logs directly through the flume terminal executable securely.
• Remote Git Integration: The Dashboard natively clones any remote https:// or ssh:// repository asynchronously without breaking boundaries.
• Mission Control Kill Switch: Immediate Docker Swarm termination routes explicitly block LLM budget-bleed spirals securely.
• Exo Auto-Discovery: Instant localized integration bridging into decentralized Macbook GPU grids locally.
• Kubernetes-Grade State Storage: Elastic & OpenBao mounts natively mapped avoiding brittle orchestration faults.
• The ReleaseFlow Matrix: Full CI/CD pipelines deploying multi-architecture binaries and analyzing CVEs natively on merge.

⚡ Quick Start (Time-to-Value)

Docker Desktop or OrbStack is mandatory. The Go CLI only orchestrates containers—zero host pollution.

# 1. Download the ecosystem
git clone https://github.com/Fremen-Labs/flume.git
cd flume

# 2. Compile and Install the CLI Native Engine 
go build -o flume cmd/flume/main.go
sudo cp ./flume /usr/local/bin/flume

# 3. Boot the Matrix
# Start the Hub-and-Spoke interactive configuration wizard
flume start 

# OR - Boot programmatically using Infrastructure-as-Code (IaC) YAML
flume start --config ./flume-mesh.yml

Note on Configuration: The interactive wizard loops infinitely, allowing you to seamlessly integrate both public Frontier API endpoints (Grok, Anthropic, Gemini) and Local hardware nodes (Ollama/Exo) on the exact same topological plane securely.

The Flume orchestrator dashboard will initialize concurrently at http://localhost:8765.

---

📚 Documentation Architecture

We follow the elite Diátaxis Documentation Framework, segmenting the Flume manuals securely into specific user intents. Please select the guide you need below:

⚙️ Operations Guides

How to handle the explicit lifecycle of the deployment safely.

• System Lifecycle (Install / Update / Destroy): Master the deterministic states of Flume natively without breaking bounds.
• Troubleshooting Matrices: Pinpoint solutions for memory mapping faults, orphaned Docker network bindings, and container delays natively.

🔮 Walkthroughs & Capabilities

How to unlock the maximum potential of Flume's native AI edge capabilities.

• GUI Onboarding & Dispatch: Zero-friction visual dashboard setup connecting active React 18 interfaces to your absolute repository safely.
• API Headless Onboarding: Explicit curl API mappings to deterministically invoke Agent planning chains outside a UI context safely.
• Zero-Cost Local Inference Setup: Bypass OpenAI completely and run open-weight models (via Exo / Ollama) through the host.docker.internal bridge smoothly.
• Zero-LLM Meta-Critic CI/CD Pipeline: Secure your repository with our mathematically exact automated garbage-collected code linter securely.
• Elastro Code Graph Indexing: Replace flawed token-RAG with AST semantic maps natively parsing variable dependencies locally.
• Inception Skills Integration: Construct deterministic, natively compiled Golang skill handlers leveraging localized Elastic Context RAG nodes without parsing latency.

📊 Observability & Metrics

How to monitor real-time gateway performance, VRAM constraints, and fallback frequency.

• Advanced Telemetry Uplink: Guide on configuring the zero-dependency Prometheus GET /metrics backend and parsing the React-based visual dashboard arrays.
• The Flume gateway exposes a zero-dependency Prometheus Text Exposition Endpoint natively at GET /metrics (port 8766 by default or internal via Docker).
• Core Insights: Tracks flume_ensemble_requests_total, flume_escalation_total, flume_ensemble_score_histogram, and internal Go heuristics (go_goroutines, go_memstats_alloc_bytes).
• Easily ingested natively by Metricbeat or raw Prometheus instances. Example Grafana queries:
  • Base Local Fallback Rate: rate(flume_escalation_total[5m])
  • Active Evaluator Models: flume_active_models
  • High VRAM Constraint Drops: increase(flume_vram_pressure_events_total[1h])

📘 Reference Materials

Deep architectural explanations of Flume's matrix integrations.

• CLI Parameter Maps: Safely inspect exactly how start, destroy, and doctor flags behave cleanly.
• The Container Matrix Architecture: Understand the strict isolation paradigm splitting the execution logic from your raw SSD safely.
• Settings & System Configuration: Deep dive into the Flume configuration GUI managing strict LLM limits, Vault keys, and GitHub/ADO tokens securely.
• Provider Architecture Interface: Detailed mapping of internal gateway translation protocols parsing Anthropic nodes and Gemini execution invariants.
• Elasticsearch Topology Map: Persistent index topology representing Flume's distributed schema patterns.