Ask
Semantic search over Hacker News, built on pgvector
Last verified:
What is Ask?
Rivestack’s ask.rivestack.io is a semantic search engine for Hacker News posts and comments. It is built to search by meaning rather than exact keywords, so users can find relevant discussions even when the wording does not match their query directly.
The site is powered by PostgreSQL with pgvector, using embeddings and HNSW vector indexing on Rivestack’s own managed infrastructure. The company says the search runs on a real Rivestack cluster and that each query hits pgvector on NVMe storage, with fast response times and production-style tuning.
The tool is aimed at people who browse Hacker News often, including developers, founders, researchers, and AI builders who want a better way to explore HN discussions. It is especially useful for semantic discovery, comparing similar topics, and surfacing relevant posts or comments without relying on exact keyword matches.
Rivestack also uses ask.rivestack.io as a live demo of its managed pgvector service. That makes the site both a search tool and a proof-of-concept for teams evaluating PostgreSQL-based semantic search for their own products.
Ask pricing
Pricing model: Freemium
The search itself is free to use. Rivestack says it has a free tier for its managed pgvector offering, with dedicated plans starting at $35 per node per month. The pricing page lists a Shared plan at $0/month for testing and prototyping, a Starter plan at $35/node/month, a Growth plan at $59/node/month, and a Scale plan at $99/node/month. Shared includes shared CPU, 256 MB RAM, pgvector enabled, 2 GB storage, about 100K vectors at 1536 dimensions, community support, and SSL. Starter includes 2 vCPU, 4 GB RAM, 55 GB NVMe storage, about 1M vectors, HNSW tuning, daily backups with 14-day PITR, a monitoring dashboard, Terraform provider, and HA readiness. Higher tiers add more compute, storage, priority support, and custom PostgreSQL configuration.
Ask pros
- Semantic search over Hacker News
- Searches by meaning, not keywords
- Covers both posts and comments
- Built on PostgreSQL and pgvector
- Uses HNSW vector indexing
- Runs on a real Rivestack cluster
- NVMe-backed search storage
- Fast query response
- Production-style vector tuning
- Useful for finding related HN discussions
- Helpful for developers and founders
- Good demo of pgvector in practice
- Shows semantic search on real data
- Works as a discovery tool for old threads
- Demonstrates managed Postgres for AI use cases
- No separate vector database required
- Simple concept for users to understand
- Free to use
Ask cons
- Limited to Hacker News content
- Not a general web search engine
- Depends on embedding quality
- May miss exact keyword-only matches
- Results can be less transparent than keyword search
- Likely less useful for very niche HN queries
- Search scope is constrained to indexed data
- No obvious advanced user filters mentioned
- Performance depends on vector model and index settings
- Not a full social/community platform
- Not suited for real-time non-HN data
Frequently asked questions about Ask
What does ask.rivestack.io do?
It is a semantic search engine for Hacker News. Instead of matching only the exact words you type, it looks for posts and comments that are similar in meaning, which helps surface relevant discussions even when the wording differs.
What content does it search?
It searches indexed Hacker News posts and comments. The site is focused specifically on HN, so it is not a general-purpose search engine for the wider web.
How is the search implemented?
The service uses PostgreSQL with the pgvector extension. Rivestack says the data is indexed with vector search technology, including HNSW, so queries are handled as nearest-neighbor semantic searches rather than keyword lookups.
Why is it fast?
Rivestack says the search runs on a real cluster with pgvector on local NVMe storage. The platform is tuned for vector workloads, and the company highlights low-latency search as one of the main benefits of the underlying infrastructure.
Do I need an account to use it?
The search is presented as free to use, and the website also says Rivestack has a free tier for people who want to try similar pgvector setups. The public HN search experience is meant to be easy to access without a heavy setup.
What is Rivestack showing with this demo?
The demo is meant to show how semantic search can run on managed PostgreSQL with pgvector. It acts as a live example of Rivestack’s infrastructure, including NVMe-backed storage, pgvector tuning, and a production-style search setup.
Who is this tool for?
It is mainly for Hacker News readers, developers, founders, and researchers who want to find related HN discussions more effectively. It is also useful for teams evaluating PostgreSQL-based semantic search for their own applications.
What are the main limitations?
The biggest limitation is that it only searches Hacker News content. Because it is semantic search, it may not always behave like exact keyword search, and its quality depends on the underlying embedding and indexing setup.
Is this the same as a vector database?
No. The site demonstrates semantic search built on PostgreSQL with pgvector, which means the vector search lives inside a standard database instead of a separate standalone vector database.
What pricing does Rivestack offer for the broader service?
Rivestack lists a $0 Shared plan for testing and prototyping, a $35/node/month Starter plan, a $59/node/month Growth plan, and a $99/node/month Scale plan. The higher plans add more compute, more storage, and more support, while the search demo itself remains free to use.