Hacktoberfest Hack Day Bengaluru x OLake

Event Description

Hacktoberfest Hack Day Bengaluru x OLake is a one-day, in-person hackathon at the OLake office in HSR Layout, focused on open-source AI.

OLake is an open-source data platform with two parts:

• OLake ingestion copies data from databases and streams (Postgres, MySQL, MongoDB, Oracle, MSSQL, DB2, Kafka, S3) into Apache Iceberg or Parquet.
• OLake Fusion keeps those Iceberg tables healthy with automated maintenance such as compaction.

This year there is only one path that you can pick and build solo:

##Ask OLake: a chat assistant for the OLake docs

Build a chat assistant for the OLake docs that answers user questions about both: setup, source configuration, sync modes, catalogs, table maintenance, deployment and troubleshooting.

Use the OLake docs and the open-source repos (olake, olake-ui, olake-helm, olake-fusion). Bring in outside sources when the docs aren't enough. Every answer must cite its sources and say clearly when it doesn't know.

It will run on a public website, so build it as if it's going to production.

What to submit (by 6:30 PM)

• A publicly reachable endpoint: POST /ask with {"question": "..."} returning {"answer": "...", "sources": ["<url>", ...]}
• A chat UI that can be embedded in a docs site
• Your repo, with a README explaining your architecture and design choices

Judging

Judges:
1. Sachin Garg, Tech Lead @ Airbnb
2. Brihati Jain, Co- Founder @ Conqur AI

Format:

All submissions are scored on the same test set after code freeze.

What we provide

A hack-day GitHub repo, with:

• Setup steps to run OLake UI locally (http://localhost:8000, login admin / password)
• Postgres seeded with small e-commerce tables (orders, customers, payments, products)
• MinIO plus an Iceberg catalog as the destination
• How to sign in and explore every API route in Swagger (http://localhost:8000/swagger/index.html)

The work, in three tiers

Core alone makes a complete entry; Stretch and Bonus are where strong entries stand out. Tool names and design are up to you.

Core: read, run and diagnose existing pipelines
- List and inspect sources, destinations and jobs, with secrets masked
- Discover the tables in a source
- Trigger a sync and report progress until it finishes
- Diagnose a failed run from its logs (summarised, not dumped)
- Pause and resume a job

Stretch: change pipelines
- Create a job: choose tables, sync mode (full_refresh, incremental, cdc) and schedule
- Configure each table on its own: sync mode, cursor field (for incremental), append or upsert, upsert type (equality deletes, position deletes, deletion vectors), partition regex, normalization and data filter. Offer only what that table supports, as returned by discovery
- Add a table or change a table's settings on an existing job. Warn and ask for confirmation when the change will clear that table's synced data
- Test and update a source's connection, and fix the broken job end to end

Bonus
- When a sync fails, suggest a concrete fix grounded in the OLake docs (for example, lower max_threads after an out-of-memory kill, or create the replication slot for Postgres CDC), and apply config fixes once the user confirms
- Destructive actions (delete a job, clear a destination) that refuse without the user's explicit confirmation
- OLake Fusion: list Iceberg tables with health details and enable maintenance

What to submit (by 6:30 PM)

• MCP server that starts with one command (docker compose up or docker run), listens on port 8080 with MCP over Streamable HTTP at /mcp, and reads OLAKE_URL, OLAKE_USERNAME and OLAKE_PASSWORD from the environment. We run every entry this way after code freeze, so this is required.
• Your repo, made public at submission with an MIT or Apache-2.0 license, and a README listing each tool, its inputs, its safety behaviour and which all tiers you completed.

Server-side safety is mandatory: mask secrets in every output and log line, and block destructive tools without an explicit confirmation.

Format

• Individual hackathon, in person. Choose Path A or Path B at check-in
• Hacking runs from 10:45 AM to the 6:30 PM code freeze
• Open to working professionals and university students
• LLM API credits provided
• Path B: bring a laptop with Docker installed (16 GB RAM recommended)

Top finalists from both paths present live to the judges.

Make your hackday smoother

• Star OLake on GitHub for easy access to the repo
• Skim the OLake docs so you know how we work
• Check out the OLake UI repo to see how it works.
• Join OLake on Slack to stay updated on events, PRs, contests, and more!

Happy hacking! #Hacktoberfest