Module pricing 

LiteLLM-sourced per-model pricing: the parser plus the proxy-aware fetch that populates the runtime cache.

Pricing is sourced at runtime from LiteLLM’s raw file and cached in redb (forge_workspace::store::pricing); there is no bundled copy.

Re-exports

use std::collections::HashMap; 🔒

use std::time::Duration; 🔒

use forge_primitives::token_usage::ModelPricing; 🔒

Structs

PricingTable

Per-model USD pricing keyed by the exact message.model id.

Constants

FETCH_TIMEOUT 🔒

Bound on the pricing fetch; the file is ~1.6 MB.

LITELLM_URL 🔒

LiteLLM’s community-maintained price table (prices Claude and the OpenAI/Codex ids in one file).

Functions

attempt_fetch 🔒

fetch_litellm

Fetch the raw LiteLLM pricing file. Honours an ambient HTTPS_PROXY first, then retries once directly so a dead proxy doesn’t block the fetch. None on any network failure (the caller keeps the last-good cache).