LiteLLM Router Map from Model Inventory (No Invented Token Counts)

Prompt

Act as a LiteLLM router architect who only uses a pasted model inventory. You write a router map the inventory already supports. You do not invent token counts, latency claims, cost totals, or success rates. This is not a live LiteLLM proxy run and not a billing audit.
You work only from Inputs. Do not invent stats, citations, quotes, URLs, names, IDs, or records that are not in Inputs.

Inputs:
- Model inventory I lock (model stubs, route cues, status notes): [Inventory]
- LiteLLM / proxy version notes I lock: [Version]
- Project or workspace label I may quote (or UNKNOWN): [Workspace]
- Required model or route names I may quote (or UNKNOWN): [SpecNames]
- Words I must not use: [Banned]
- What I must never invent (token counts, latency claims, cost totals, success rates): [Never]
- Output format: [Format]
- Language: [Lang]

Generate:
1. Honesty ledger: Inventory nouns, Version, Workspace, SpecNames, Lang. Forbidden: invented token counts, latency claims, cost totals, success rates. Banner: not a live LiteLLM proxy run; not a billing audit.
2. Router map table: one row per Inventory model stub or route cue. Missing status notes write NOT IN INPUTS. Use LiteLLM model_list, Router, fallbacks, rpm/tpm limits, and litellm_params language when Inventory supports it. Never print token-count VALUES not in Inventory.
3. Spec name set: only names in SpecNames. Unnamed models stay NOT IN INPUTS. Never print latency or cost VALUES not in Inventory.
4. Version lock: print Version. Refuse LiteLLM features newer than Version if Version is named (no unreleased Guardrails AI claims).
5. Refuse list: inventing token counts, inventing latency claims, inventing cost totals, inventing success rates.
6. Compliance pass: quote Banned and Never hits. Cut them. Format as Format.

Constraints:
- Map from Inventory only. No invented token-count VALUES. Teach LiteLLM router mapping, not a generic proxy swap.
- 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.

LiteLLM Router Map from Model Inventory (No Invented Token Counts) - Result

Examples

Example Input

Inventory: model Harbor Quay berth_chat as pasted route cue gpt-4o-mini as pasted status note draft as pasted; model River Ops lock_reason as pasted route cue claude-3-5-haiku as pasted. Token count NOT IN INPUTS. Latency NONE.
Version: LiteLLM 1.40.0 as pasted (do not invent unreleased Guardrails AI APIs)
Workspace: Harbor Quay Harbor Proxy as pasted
SpecNames: berth_chat as pasted; lock_reason as pasted. tide_embed UNKNOWN.
Banned: guaranteed token count, latency invent, cost invent, success invent
Never: invent token counts, invent latency claims, invent cost totals, invent success rates
Format: ledger + table + specs + version + refuse + compliance
Lang: English

Example Output

1. Ledger. Inventory: model Harbor Quay berth_chat route cue gpt-4o-mini status note draft; model River Ops lock_reason route cue claude-3-5-haiku. Token count NOT IN INPUTS. Latency NONE. Version LiteLLM 1.40.0. Workspace Harbor Quay Harbor Proxy. SpecNames berth_chat; lock_reason; tide_embed UNKNOWN. Lang English. Banner: not a live LiteLLM proxy run; not a billing audit. Forbidden: invented token counts, latency claims, cost totals, success rates, guaranteed token count, latency invent, cost invent, success invent.

2. Router map.
1. Harbor Quay berth_chat | route gpt-4o-mini as pasted | status draft as pasted | tokens NOT IN INPUTS | latency NONE | LiteLLM model_list / Router map only. Token-count VALUES not printed.
2. River Ops lock_reason | route claude-3-5-haiku as pasted | status NOT IN INPUTS | tokens NOT IN INPUTS | latency NONE.
Cost totals not printed. Success rates not invented.

3. Spec name set. berth_chat; lock_reason as SpecNames. tide_embed UNKNOWN so write tide_embed NOT IN INPUTS. No token-count VALUES printed. No third model invented.

4. Version lock. LiteLLM 1.40.0 as pasted. Unreleased Guardrails AI APIs not used. fallbacks schema NOT IN INPUTS.

5. Refuse. Token invent: refused. Latency invent: refused. Cost invent: refused. Success invent: refused. Guaranteed token count: refused.

6. Compliance. Banned hits none. Format ledger+table+specs+version+refuse+compliance. Gaps: status for River Ops lock_reason, tide_embed decision, rpm limit, api_base if any, litellm_params 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.