Helicone Request Cost Ledger from Trace Notes (No Invented Token Totals)

Prompt

Act as a Helicone LLM observability engineer who only uses pasted trace notes. You compile a request cost ledger the notes already support. You do not invent token totals, latency percentiles, monthly spend ranks, or cache hit rates. This is not a live Helicone dashboard pull and not a FinOps billing estimator.
You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs.

Inputs:
- Trace notes I lock (request labels, model cues, property cues): [TraceNotes]
- Helicone plan or SDK notes I lock: [Version]
- Project or org label I may quote (or UNKNOWN): [ProjectLabel]
- Model name cues already present (or UNKNOWN): [ModelCues]
- Property or user cues already present (or UNKNOWN): [PropertyCues]
- Request path or endpoint cues already present (or UNKNOWN): [PathCues]
- Cost field cues already present (or UNKNOWN): [CostCues]
- Words I must not use: [Banned]
- What I must never invent (token totals, latency percentiles, spend ranks, cache hit rates): [Never]
- Output format: [Format]
- Language: [Lang]

Generate:
1. Honesty ledger: TraceNotes nouns, Version, ProjectLabel, ModelCues, PropertyCues, PathCues, CostCues, Lang. Forbidden: invented token totals, latency percentiles, spend ranks, cache hit rates.
2. Cost ledger checklist: one checkbox row per TraceNotes request label. Attach only ModelCues and CostCues named beside that request. Missing cost write NOT IN INPUTS.
3. Model map table: for each ModelCues entry, list request labels that name it. Do not invent a 2.1M token total if absent.
4. Property caution block: quote PropertyCues only. Auto-budget alert packs not in TraceNotes stay NOT IN INPUTS.
5. Refuse list: inventing 2.1M token totals, inventing p95 840ms latency, inventing spend rank #1, inventing 97% cache hit rates.
6. Compliance pass: quote Banned and Never hits. Cut them. Print request and model counts from TraceNotes only. Format as Format.

Constraints:
- Request cost ledger from TraceNotes only. No invented token totals.
- Honor Version. No emojis. Not a live Helicone requests console.

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.

Helicone Request Cost Ledger from Trace Notes (No Invented Token Totals) - Result

Examples

Example Input

TraceNotes: request Harbor Quay chat as pasted model gpt-4o-mini as pasted cost cue prompt_tokens field as pasted; request Midpoint Embed as pasted model text-embedding-3-small as pasted. Token total invent NONE. Latency invent NONE.
Version: Helicone TypeScript SDK as pasted (do not invent unreleased AI cost coach)
ProjectLabel: Harbor Quay LLM gateway as pasted
ModelCues: gpt-4o-mini as pasted; text-embedding-3-small as pasted. Cedar claude-3 UNKNOWN.
PropertyCues: property user_id hashed as pasted for Harbor Quay chat. Midpoint Embed property UNKNOWN.
PathCues: path /v1/chat/completions as pasted for Harbor Quay chat. Midpoint Embed path UNKNOWN.
CostCues: prompt_tokens field as pasted for Harbor Quay chat; Midpoint Embed cost field NOT IN INPUTS.
Banned: 2.1M tokens, p95 840ms, guaranteed spend rank #1
Never: invent token totals, invent latency percentiles, invent spend ranks, invent cache hit rates
Format: ledger + cost checklist + model map + property caution + refuse + compliance
Lang: English

Example Output

1. Ledger. TraceNotes: request Harbor Quay chat + gpt-4o-mini + prompt_tokens; request Midpoint Embed + text-embedding-3-small. Token total invent NONE. Latency invent NONE. Version Helicone TypeScript SDK. ProjectLabel Harbor Quay LLM gateway. ModelCues gpt-4o-mini; text-embedding-3-small. PropertyCues user_id hashed for Harbor Quay chat; Midpoint Embed property UNKNOWN. PathCues /v1/chat/completions for Harbor Quay chat; Midpoint Embed path UNKNOWN. CostCues prompt_tokens for Harbor Quay chat; Midpoint Embed cost field NOT IN INPUTS. Lang English. Forbidden: invented token totals, latency percentiles, spend ranks, cache hit rates, 2.1M tokens, p95 840ms, guaranteed spend rank #1.

2. Cost ledger checklist.
[ ] Harbor Quay chat | model gpt-4o-mini as pasted | cost cue prompt_tokens as pasted.
[ ] Midpoint Embed | model text-embedding-3-small as pasted | cost field NOT IN INPUTS.
Cedar claude-3 not listed.

3. Model map table.
gpt-4o-mini | request Harbor Quay chat as pasted | token total NOT IN INPUTS so do not invent 2.1M.
text-embedding-3-small | request Midpoint Embed as pasted.
Second gpt-4o-mini row not invented.

4. Property caution. user_id hashed as pasted for Harbor Quay chat. Auto-budget alert pack NOT IN INPUTS. Do not invent cache-hit-rate packs.

5. Refuse. 2.1M token totals: refused. p95 840ms latency: refused. spend rank #1: refused. 97% cache hit rates: refused. Unreleased AI cost coach: refused.

6. Compliance. Banned hits none. Requests 2. Models 2. Format ledger+cost checklist+model map+property caution+refuse+compliance. Gaps: Midpoint Embed cost field, Midpoint Embed property, Midpoint Embed path, Cedar claude-3, token total.

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.