Intent-first Text2Cypher: LLM never writes Cypher, but only 1 of 4 runs correct
JeremyCMorgan · x · 2026-10-28
The Ontographia project's part-2 article details an architecture for natural-language querying of Neo4j: the LLM's only job is extracting a structured Intent (start node, relationship paths, filters, returns), which a fixed core validates against the ontology and deterministically compiles into parameterized Cypher — nothing the LLM writes reaches the database as query text.
The honest benchmark: only 1 of 4 end-to-end attempts was fully correct. Syntax is solved; semantics is not — intent extraction remains the bottleneck. Open source at github.com/edgesentry/ontographia, with stress tests and public-dataset validation coming in part 3.
More from coding & agent
- One invisible space makes Claude Code and other LLM graders flip verdicts — maverick_man1111 · 2026-10-28
- Open-source VoiceStudio offers local voice cloning and dubbing in 646 languages, free ElevenLabs rival — thisdudelikesAI · 2026-10-28
- cellarmcp: a remote MCP server with OAuth 2.1 for managing your wine cellar — Emu-Outside · 2026-10-28
- MCP Python SDK maintainer wins Agentic AI Foundation Outstanding Technical Contribution Award — samuelcolvin · 2026-10-28
- Maestro adds GitHub Label Added trigger that spins up agent pipelines — Dave_Maynor · 2026-10-28
- Agents Don't Call Tools: The LLM Only Generates Text You Parse and Execute — ariG23498 · 2026-10-28