🤖 AI Tools
LangSmith Run Trace Filter Checklist from Ops Notes (No Invented Latency Ms)
Compile a LangSmith run trace-filter checklist from pasted ops notes only. No invented latency ms, error ranks, or eval scoreboards. Not a live LangSmith sync.
0Reviews
Prompt
Act as a LangSmith run trace-filter checklist operator who only uses pasted ops notes. You compile a run trace-filter checklist the notes already support. You do not invent latency ms, error ranks, eval scoreboards, or token cost claims. This is not a live LangSmith sync, not a Helicone invent, and not ML ops consulting advice. You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs. Inputs: - Ops notes I lock (run stubs, trace cues, filter fragments): [OpsNotes] - LangSmith project or workspace notes I lock: [Version] - Project or env label I may quote (or UNKNOWN): [ProjectLabel] - Run titles already present (or UNKNOWN): [RunTitles] - Trace cues already present (or UNKNOWN): [TraceCues] - Filter cues already present (or UNKNOWN): [FilterCues] - Tag cues already present (or UNKNOWN): [TagCues] - Words I must not use: [Banned] - What I must never invent (latency ms, error ranks, eval scoreboards, token cost claims): [Never] - Output format: [Format] - Language: [Lang] Generate: 1. Honesty ledger: OpsNotes nouns, Version, ProjectLabel, RunTitles, TraceCues, FilterCues, TagCues, Lang. Banner: not ML ops consulting advice; not a live LangSmith sync. Forbidden: invented latency ms, error ranks, eval scoreboards, token cost claims. 2. Run trace-filter checklist: one checkbox row per RunTitles entry. Attach only TraceCues named beside that run in OpsNotes. Missing trace write NOT IN INPUTS. 3. Filter sketch: for each FilterCues entry, list runs that name it. Do not invent a 420 ms latency claim if absent. 4. Tag caution block: quote TagCues only. Dataset packs not in OpsNotes stay NOT IN INPUTS. 5. Refuse list: inventing 420 ms latency, inventing error ranks, inventing eval scoreboards, inventing token cost claims. 6. Compliance pass: quote Banned and Never hits. Cut them. Print run and trace counts from OpsNotes only. Format as Format. Constraints: - Run trace-filter checklist from OpsNotes only. No invented latency ms. - Honor Version. No emojis. Not a live LangSmith dashboard. Not ML ops consulting advice.
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
OpsNotes: run title Harbor Quay Chat Turn as pasted trace cue retriever as pasted filter cue error-true as pasted; run title Quay Storm Tool Call as pasted trace cue tool-node as pasted. Latency invent NONE. Error invent NONE. Version: LangSmith as pasted (do not invent unreleased AI latency coach) ProjectLabel: Cedar Pier prod project as pasted RunTitles: Harbor Quay Chat Turn as pasted; Quay Storm Tool Call as pasted. Batch pack UNKNOWN. TraceCues: trace cue retriever as pasted for Harbor Quay Chat Turn; trace cue tool-node as pasted for Quay Storm Tool Call. Feedback pack UNKNOWN. FilterCues: filter cue error-true as pasted for Harbor Quay Chat Turn; Quay Storm Tool Call filter NOT IN INPUTS. TagCues: tag cue staging as pasted for Harbor Quay Chat Turn. Dataset pack UNKNOWN. Banned: 420 ms, error rank #1, guaranteed eval scoreboard Never: invent latency ms, invent error ranks, invent eval scoreboards, invent token cost claims Format: ledger + run trace-filter checklist + filter sketch + tag caution + refuse + compliance Lang: English
Example Output
1. Ledger. OpsNotes: run title Harbor Quay Chat Turn + trace cue retriever + filter cue error-true; run title Quay Storm Tool Call + trace cue tool-node. Latency invent NONE. Error invent NONE. Version LangSmith. ProjectLabel Cedar Pier prod project. RunTitles Harbor Quay Chat Turn; Quay Storm Tool Call. TraceCues retriever for Harbor Quay Chat Turn; tool-node for Quay Storm Tool Call. FilterCues error-true for Harbor Quay Chat Turn; Quay Storm Tool Call filter NOT IN INPUTS. TagCues staging for Harbor Quay Chat Turn. Dataset pack UNKNOWN. Lang English. Banner: not ML ops consulting advice; not a live LangSmith sync. Forbidden: invented latency ms, error ranks, eval scoreboards, token cost claims, 420 ms, error rank #1, guaranteed eval scoreboard. 2. Run trace-filter checklist. [ ] Harbor Quay Chat Turn | trace retriever as pasted. [ ] Quay Storm Tool Call | trace tool-node as pasted. Feedback pack not attached. Batch pack not added. 3. Filter sketch. filter cue error-true | run Harbor Quay Chat Turn as pasted. Quay Storm Tool Call filter | NOT IN INPUTS. Latency ms NOT IN INPUTS so do not invent 420 ms. Second filter cue not invented. 4. Tag caution. tag cue staging as pasted for Harbor Quay Chat Turn. Dataset pack UNKNOWN. Do not invent token cost claim packs. 5. Refuse. 420 ms latency: refused. error ranks: refused. eval scoreboards: refused. token cost claims: refused. Unreleased AI latency coach: refused. 6. Compliance. Banned hits none. Runs 2. Traces named 2. Format ledger+run trace-filter checklist+filter sketch+tag caution+refuse+compliance. Gaps: Quay Storm Tool Call filter, Feedback pack, Batch pack, Dataset pack, latency ms. 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.