EchoPipeline-AI

What it does

EchoPipeline-AI is an ambient RevOps intelligence bridge that captures voice-first sales streams via Amazon Alexa+, processes lead batches using Model Context Protocol (MCP) Streamable HTTP endpoints, and performs autonomous multi-step reasoning using Gemini 3.6 Flash.

Every ingested cycle calculates pipeline lead deltas, qualifies risk using the RIGS matrix, settles escrow trust metrics, and dispatches real-time structured Slack alerts while committing state changes to PostgreSQL.

How we built it

• Backend Architecture: Engineered with Python 3.11 and FastAPI, featuring asynchronous connection pooling via asyncpg.

• Autonomous Reasoning: Integrated live multi-step Gemini 3.6 Flash evaluation cycles to dynamically parse intent strings and score lead batches.

• Protocol & Interfaces: Implemented the MCP Streamable HTTP specification alongside rich Slack Block Kit notifications and a live HTML5 dashboard.

• Infrastructure: Containerized with Docker and continuously deployed on Render.

Challenges we ran into

• Balancing real-time reasoning latency with strict zero-latency ambient response thresholds.

• Ensuring full auditability and state recovery when network or third-party API keys are missing by engineering fallback local synthesis matrices.

• Structuring rich, mobile-readable Slack block payloads to convey dynamic execution steps and financial trust deltas seamlessly.

• Production uptime, migrated from railway to render, due to subscription expiration date, to keep EchoPipeline-AI online during the Hackathon judging phase.

Accomplishments that we're proud of

• Achieved zero-latency telemetry execution with full multi-step reasoning visibility.

• Built end-to-end integration across voice input, local PostgreSQL persistence, Gemini 3.6 flash autonomous evaluation, and Slack broadcasting.

• Verified production execution under high-volume multi-lead test batches.

What we learned

• Best practices for streaming asynchronous MCP JSON-RPC endpoints within serverless containerized environments.

• Effective prompt-engineering strategies for dynamic, structured step-by-step reasoning outputs tailored for executive telemetry dashboards.

What's next for EchoPipeline-AI

• Implementing full bidirectional voice execution responses directly back to Echo Studio hardware.

• Adding multi-agent consensus protocols where Grok, Gemini, and Claude collaborate on high-value escrow settlements before committing DB updates.

Built With

• alexa
• asyncpg
• docker
• fastapi
• gemini
• genai-sdk
• httpx
• mcp
• postgresql
• python
• railway
• slack-api