datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
synthetic-hvac-engineering-precheck
synthetic-hvac-engineering-precheck (v0.1)
Scope
Programmatically generated Q/A pairs for building-services quick checks (HVAC/hydronics). Focus is on:
unit hygiene (explicit conversions)
deterministic equations and step-by-step workflows
basic sanity checks and “missing inputs” prompts
This is pre-design / pre-check material. It is not a substitute for code-compliant design or professional sign-off.
Data
Each row is a JSON object with a single field:
text:… See the full description on the dataset page: https://huggingface.co/datasets/Irfanuruchi/synthetic-hvac-engineering-precheck.hvac-instruct-v2-sft
hvac-instruct-v2-sft
Hand-written SFT pairs for the v2 HVAC field assistant fine-tune.
83 conversations, multi-turn included
Format: OpenAI-style messages array per row
System prompt includes retrieved manual chunks (same shape as inference pipeline lib/chat/answer.js)
Five capabilities: slang_to_query, yes_no_advance, terse_cited, tier_2_broad, oos_refusal
Indoor wired-remote error codes only (01, 02, 03, 04, 05, 06, 09); CH-codes out of scope
Blocklisted against… See the full description on the dataset page: https://huggingface.co/datasets/vishnu-samavedula/hvac-instruct-v2-sft.