March 17, 2026
Trench coat or true fusion?
Show HN: Antfly: Distributed, Multimodal Search and Memory and Graphs in Go
One engine to search it all—or three tools in a trench coat?
TLDR: Antfly claims one engine that searches by keywords, meaning, and relationships across text, images, audio, and video. Commenters are split between excitement and “prove it,” grilling how the hybrid ranking really works and flagging a broken SDK link—because if it’s real, it could replace three separate tools.
Antfly just hit the scene promising a “do‑it‑all” search brain: one box that can find stuff by keywords, by meaning (AI), and by connections between things (graphs), across text, images, audio, and video—with built‑in chatbots powered by retrieval‑augmented generation (RAG). The community reaction? A loud mix of wow and wait, show us.
The top vibe is healthy skepticism. User thefogman asks if this is truly one unified query or “three separate indexes that happen to live in the same binary.” Translation: Is this magic, or a trench coat with three tools inside? Another dev, didip, goes straight for receipts, poking at tests and demanding to see how hybrid ranking actually merges “semantic and keyword” results and reorders them. On the other side, hype is real—jnstrdm05 drops a simple “This looks sick!”, while maintainer kingcauchy jumps in with confident energy, talking migrations and a “linear merge API” to bring your data over without tears.
Then there’s drama of the everyday kind: Linell spots a broken TypeScript SDK link returning 404, prompting classic “am I doing it wrong?” energy. Meanwhile, Antfly’s “Antfarm” dashboard and promise of plug‑and‑play models has people curious, but they want proof—demos, docs, and working links. If Antfly’s fusion actually works, devs might finally ditch juggling three systems. Until then, the crowd’s chanting: ship it, show it, prove it. Check the docs at antfly.io/docs.
Key Points
- •Antfly is a distributed, multimodal search engine in Go that combines full‑text (BM25), dense/sparse vectors (SPLADE), and graph traversal in one system.
- •It includes built‑in RAG agents, automatic embeddings/chunking/graph extraction, and a web dashboard (Antfarm) for search, RAG, and knowledge graphs.
- •Deployment is simple with a single‑node swarm mode or Docker; Termite runs ML inference locally by default.
- •Operational features include shard‑level ACID transactions, document TTL, object‑storage support (S3/MinIO/R2), SIMD/SME acceleration, and Raft‑based sharding/replication.
- •Ecosystem includes SDKs (Go, TS, Python), React UI components, a PostgreSQL extension (pgaf), ingestion/eval tools, a Kubernetes operator, MCP server, and A2A protocol support.