feat(lens): estimate OTEL costs with shared Rust pricing

deploy/lens/Dockerfile

@@ -6,6 +6,8 @@ RUN apk add --no-cache rust build-base cmake perl pkgconf openssl-dev libseccomp
 WORKDIR /src
 COPY .cargo/ .cargo/
 COPY litellm-rust/ litellm-rust/
+COPY model_prices_and_context_window.json model_prices_and_context_window.json
+COPY litellm/integrations/otel/model/providers.json litellm/integrations/otel/model/providers.json
 COPY litellm/proxy/lens/prompts/ litellm/proxy/lens/prompts/
 WORKDIR /src/litellm-rust
 ENV CARGO_PROFILE_RELEASE_DEBUG=0 CARGO_PROFILE_RELEASE_STRIP=symbols

deploy/lens/Dockerfile.dockerignore

@@ -13,3 +13,9 @@
 !litellm-rust/
 !litellm-rust/**
 litellm-rust/target/
+
+!model_prices_and_context_window.json
+!litellm/integrations/
+!litellm/integrations/otel/
+!litellm/integrations/otel/model/
+!litellm/integrations/otel/model/providers.json

litellm-rust/Cargo.toml

@@ -9,6 +9,8 @@ license = "MIT"
 repository = "https://github.com/BerriAI/litellm"
 
 [workspace.dependencies]
+litellm-cost = { path = "crates/cost" }
+litellm-model-catalog = { path = "crates/model-catalog" }
 litellm-config = { path = "crates/config" }
 litellm-router = { path = "crates/router" }
 litellm-tracing = { path = "crates/tracing" }

litellm-rust/crates/cost/Cargo.toml

@@ -5,6 +5,11 @@ edition.workspace = true
 license.workspace = true
 repository.workspace = true
 
+[dependencies]
+serde_json.workspace = true
+chrono = { version = "0.4", default-features = false, features = ["std"] }
+chrono-tz = "0.10.4"
+
 [dev-dependencies]
 criterion.workspace = true
 proptest.workspace = true

litellm-rust/crates/cost/README.md

@@ -1,13 +1,17 @@
 # litellm-cost
 
-This crate calculates text token charges from rates and usage supplied by its caller. It is standalone and has no Python bridge or proxy integration
+This crate calculates text token charges from rates and usage supplied by its caller. It has no Python dependency and is used by standalone Lens for display estimates
 
 Call `compile(&pricing)` once for an immutable plan, then `plan.calculate(&request)` for each supported request. `calculate(&pricing, &request)` compiles on each call. A successful result exposes pre-multiplier component costs, selected rates, the multiplier, and derived `input()`, `output()`, and `total()` values
 
 The caller states whether `prompt_tokens` includes cache tokens. Threshold selection uses total input tokens for either convention and selects one rate for the whole request. Thresholds are sorted when compiled, and duplicate thresholds or tier overrides fail deterministically. `Fast` selects priority rates; unknown tiers use standard rates
 
-`Rate::Missing`, `Rate::Null`, and `Rate::Value(0.0)` remain distinct. Missing cache rates fall back to the selected input rate, and an absent one-hour write rate falls back to the selected write rate. Missing input or output rates return typed errors, including for zero usage. Python's sparse-entry behavior remains outside this native contract
+`Rate::Missing`, `Rate::Null`, and `Rate::Value(0.0)` remain distinct. Missing cache rates fall back to the selected input rate, and an absent one-hour write rate falls back to the selected write rate. Missing input or output rates return typed errors, including for zero usage. These supplied-rate API semantics also remain unchanged when Lens uses catalog estimation
 
 The supported off-peak shape is one non-wrapping UTC daily window. The caller supplies the applicable regional multiplier after provider-specific selection. Negative or non-finite rates, ambiguous rules, inconsistent cache counts, incomplete write splits, invalid windows and overflow return errors. Callers must decline unsupported inputs before native execution if their public contract accepts those shapes
 
-This crate does not select models, read catalogs, fetch provider prices, normalize multimodal usage, process provider-reported costs, or calculate non-token charges. It does not change proxy behavior. The reference fixture was generated by `tests/generate_python_reference.py` against the Python implementation at the commit recorded in `tests/python_reference.tsv`, using synthetic rates and fixed usage
+`catalog::estimate` selects effective rates from a supplied raw catalog row. It supports structured tiers, service tiers, provider thresholds and cache defaults, inclusive reasoning output, and captured-time off-peak schedules with weekday timezones. It validates selected input/output rates even at zero usage, and cache/reasoning rates only when they affect the amount. Missing context or invalid applicable prices return `None`; explicit zero remains a known price. Both APIs use the same component arithmetic
+
+This crate does not select models, fetch provider prices, normalize multimodal usage, process provider-reported costs, or calculate non-token charges. `litellm-model-catalog` owns offline identity and alias resolution, and `litellm-traces` owns GenAI usage admission. Lens reads `supports_token_only_pricing: false` from the resolved catalog row to decline models whose costs cannot be represented by ordinary token charges. Set this flag to false for new dedicated search, agentic or router models, including separately hosted catalog entries; model names never determine eligibility. The same flag is exposed through Python `get_model_info`. Missing, null or true retains ordinary token eligibility; malformed values remain unknown. Known requested-model requirements are preserved when the served model differs. Uncovered request fees and reported search, modality, citation, computer-use and code-execution usage also remain unknown. Optional search capability alone does not disable ordinary token estimates
+
+Lens bundles the repository catalog into its release, so its labeled estimates use deployed catalog prices. Matched gateway spend remains authoritative. The reference fixture was generated by `tests/generate_python_reference.py` against the Python implementation at the commit recorded in `tests/python_reference.tsv`, using synthetic rates and fixed usage