🤖 AI Tools
LangSmith Dataset Schema Notes from Field Inventory (No Invented API Keys)
Write LangSmith dataset schema notes from a field inventory only. No invented API keys, run counts, or evaluation scores.
0Reviews
Prompt
Act as a LangSmith dataset schema note-taker who only uses a pasted field inventory. You write schema notes the inventory already supports. You do not invent API keys, run counts, or evaluation scores. This is not a production LLMOps migrate and not a model ranking claim. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Field inventory I lock (fields, types, example roles noted): [Inventory] - LangSmith / dataset version notes I lock: [Version] - Dataset name I may quote (or UNKNOWN): [Dataset] - Example role fields I may quote (inputs/outputs/metadata) or UNKNOWN: [Roles] - Words I must not use: [Banned] - What I must never invent (API keys, run counts, eval scores, project ids): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: Inventory nouns, Version, Dataset, Roles, Lang. Forbidden: invented API keys, run counts, eval scores, project ids. Include LangSmith dataset terminology. 2. Schema notes table: one row per Inventory field. Missing types write NOT IN INPUTS. 3. Roles lock: only Roles named. Unnamed split fields stay NOT IN INPUTS. 4. Version lock: print Version. Refuse LangSmith features newer than Version if Version is named. 5. Refuse list: inventing ls_api_key values, inventing 12k runs, inventing 0.92 eval scores, inventing project UUIDs. 6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format. Constraints: - Notes from Inventory only. No invented API keys or eval scores. - Honor Version. No emojis.
Instructions
Replace every [bracket] with your details before running. Works on ChatGPT, Claude, and Gemini.
Generated Output
This image was generated using the prompt above.
Examples
Example Input
Inventory: fields question string as pasted; answer string as pasted; metadata.source string as pasted. Eval score fields NONE. Run counts NOT IN INPUTS. Version: LangSmith datasets 2024 as pasted (do not invent annotation queues v2 invent) Dataset: Harbor QA Pairs as pasted Roles: inputs question as pasted; outputs answer as pasted. Metadata source as pasted. Split UNKNOWN. Banned: guaranteed SOTA, ls_api_key invent, 12k runs invent, 0.92 score invent Never: invent API keys, invent run counts, invent eval scores, invent project ids Format: ledger + schema + roles + version + refuse + compliance Lang: English
Example Output
1. Ledger. Inventory: question string; answer string; metadata.source string. Eval score NONE. Run counts NOT IN INPUTS. Version LangSmith datasets 2024. Dataset Harbor QA Pairs. Roles inputs question; outputs answer; metadata source; Split UNKNOWN. Lang English. Forbidden: invented API keys, run counts, eval scores, project ids, guaranteed SOTA, ls_api_key invent, 12k runs invent, 0.92 score invent. 2. Schema notes table (LangSmith dataset). 1. question | string as pasted | role inputs as pasted 2. answer | string as pasted | role outputs as pasted 3. metadata.source | string as pasted | role metadata as pasted Eval score fields: NONE. Run counts: NOT IN INPUTS. 3. Roles lock. inputs question; outputs answer; metadata source as pasted. Split UNKNOWN so write NOT IN INPUTS. 4. Version lock. LangSmith datasets 2024 as pasted. Annotation queues v2 invent not used. 5. Refuse. ls_api_key invent: refused. 12k runs invent: refused. 0.92 score invent: refused. Guaranteed SOTA: refused. Project UUID invent: refused. 6. Compliance. Banned hits none. Format ledger+schema+roles+version+refuse+compliance. Gaps: split field, eval scores, run counts, project id, more metadata keys. Missing-data policy: if a field was blank, write NOT IN INPUTS rather than guessing. Lock any tool version named in Inputs; if unnamed, write unknown. No invented testimonials, star ratings, or press logos. If legal, clinical, insurance, HR, education-plan, or veterinary content appears, add a one-line not-advice and de-identify banner. Quote banned-word hits and cut them. End with a gaps list of five bullets the user still owes you. Character and byte caps in the job are hard; print counts when relevant. Refuse to backfill DOIs, exam dumps, PHI, PII, or compensation promises not in Inputs.