This is a submission for the Agentic Postgres Challenge with Tiger Data
🧠 What I Built
MindsEye Agentic is an experimental, time-aware “cognitive event tracker” built on Agentic Postgres (Tiger Cloud).
It acts as a thinking layer for real-time systems — ingesting, timestamping, and analyzing events with time-labeled metadata.
Each event is a micro-signal (a moment of “cognition”) that can later be searched, bucketed, and visualized as part of a live memory flow.
The concept stems from my broader MindsEye research at SageWorks AI, exploring how AI systems can “see” timelines — not just data.
This version focuses on:
Recording time-labeled events into a hypertable
Aggregating those events via time_bucket()
Searching them semantically using PostgreSQL’s pg_trgm similarity
Laying the foundation for future GPT-powered “Agents of Cognition”
🚀 Demo & Repository
Although we couldn’t deploy the live API due to budget limits on hosting, the full system is ready and connected to Tiger Cloud — schema, routes, and data flow are complete.
GitHub Repository: PEACEBINFLOW/mindseye-agentic
Tiger Cloud: connected via DATABASE_URL (Timescale + Trigram enabled)
API Endpoints:
GET /health — DB status check
POST /events — add time-labeled events
GET /events — retrieve recent events
GET /events/stats — rollups via time_bucket()
GET /events/search — trigram similarity search
The repo contains:
server/
src/
index.js # Express API + Tiger connection
pg.js # PostgreSQL pool with SSL
client/
index.html # Frontend placeholder
vite.config.js # React/Vite setup for charts
All database logic is live and verified via Tiger’s SQL Editor — hypertable creation, indexes, and mock event seeding.
⚙️ How I Used Agentic Postgres
MindsEye Agentic makes use of several Agentic Postgres capabilities:
Timescale hypertables for event time-bucketing and high-frequency insert performance.
pg_trgm extension to enable fuzzy/semantic text search over event descriptions.
Fluid Storage (through Tiger Cloud’s managed scaling) to handle continuous event writes.
MCP agentic concept is reflected in the design: each API call acts as a “memory agent,” inserting or interpreting time-labeled signals.
💡 Overall Experience
Building on Tiger Cloud was incredibly fluid — the PostgreSQL base with built-in AI-oriented extensions (Timescale + Trigram) made the system design simple but powerful.
I was impressed by how Postgres suddenly felt like an AI database once hypertables and similarity search were combined.
Challenges & Learnings
Deployment costs: I couldn’t finalize live hosting on Render due to resource limits, but the backend runs perfectly in local and containerized setups.
SSL enforcement: Tiger Cloud requires ssl: { rejectUnauthorized: false } for Node’s pg library — a small but important detail.
Insight: Postgres, when augmented with Timescale + Trigram, starts to behave like a cognitive log — perfect for real-time “agentic” experiments.
🧩 What’s Next
Add a GPT-driven summarizer endpoint (/events/insights) to auto-describe spikes and anomalies.
Integrate Tiger CLI to run multi-agent forks comparing model runs over the same dataset.
Build a web dashboard to visualize “event flows” as living neural timelines.
Built by Peace Thabiwa
— Founder, SageWorks AI (Botswana).
Exploring AI that thinks in timelines — not just data rows.
https://github.com/PEACEBINFLOW/mindseye-agentic/tree/main
Top comments (0)