feat: Async Rust OCR Bridge and MCP OAuth UI Restore

litellm-rust/ADDING_A_PROVIDER.md

@@ -1,9 +1,10 @@
 # Adding a provider / route to litellm-rust
 
-Three layers, same for every route (see `ocr` and `realtime` as references):
-
-1. **Transform contract (pure)** — `crates/core/src/<route>/transformation.rs`: a `…ProviderConfig` trait (URL build + request/response transforms) + types in `types.rs`. No network, env, or auth.
-2. **Provider config (pure)** — `crates/providers/src/<provider>/<route>/transformation.rs`: implement that trait as a `const <PROVIDER>_<ROUTE>_CONFIG`, mirroring the Python provider tree. Add parity unit tests.
-3. **HTTP / transport (the host)** — `crates/providers/src/<route>.rs` (e.g. `ocr.rs`, `realtime.rs`): the callable fn (`run_ocr`, `realtime`). It resolves the key, builds the auth header, builds URL + transforms via the config, then does the network call. This is the only layer allowed to do I/O.
-
-**Calling:** the host invokes the route fn — the Python bridge calls `run_ocr`; the `ai-gateway` server calls `realtime`. Register new modules in `lib.rs` / `mod.rs`, then run `cargo fmt && cargo clippy --workspace -- -D warnings && cargo test --workspace`.
+- Keep the route contract pure in `crates/core/src/<route>/`: define the typed request/response structs and a provider config trait with no network, env, auth, or logging.
+- Add provider identity to the repo-root `provider_endpoints_support.json`: use the LiteLLM provider slug, display name, docs URL, and endpoint support flags. Put optional stable provider-level base URL defaults under the top-level `default_creds` map; keep route-specific API key env vars in provider config/transform code so key resolution has one owner.
+- Put provider-specific transforms in `crates/providers/src/<provider>/<route>/transformation.rs`, mirroring the Python provider tree and exposing a `const <PROVIDER>_<ROUTE>_CONFIG`.
+- The provider config owns three pure steps: map LiteLLM params, transform the LiteLLM request into the provider request, and transform the provider response back into the LiteLLM response.
+- If the provider has a reverse or normalization step, keep it pure and explicit next to the transforms; do not hide reverse mapping inside the HTTP transport.
+- Route host functions in `crates/providers/src/<route>.rs` must be async: resolve auth/base URL, call the transforms, send with async transport, then call the response transform. Use Tokio/async all the way through Rust route I/O; only the PyO3 sync compatibility wrapper should `block_on` the async route, and it must release the GIL while waiting.
+- Do not add per-provider HTTP clients casually. Today Rust cannot call Python's `BaseLLMHTTPHandler`; if a route needs end-to-end Rust I/O, keep the async transport route-scoped, opt-in from Python, and do not broaden it to more providers until there is a shared Rust HTTP abstraction.
+- Register modules in `lib.rs` / `mod.rs`, add parity tests for params/request/response behavior, then run `cargo fmt && cargo clippy --workspace --all-targets --locked -- -D warnings && cargo test --workspace --locked`.

litellm-rust/Cargo.toml

@@ -15,7 +15,8 @@ repository = "https://github.com/BerriAI/litellm"
 litellm-core = { path = "crates/core" }
 litellm-providers = { path = "crates/providers" }
 pyo3 = "0.23.5"
-reqwest = { version = "0.12", default-features = false, features = ["blocking", "json", "rustls-tls"] }
+pyo3-async-runtimes = { version = "0.23.0", features = ["tokio-runtime"] }
+reqwest = { version = "0.12", default-features = false, features = ["json", "rustls-tls", "http2"] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
 thiserror = "2.0"

litellm-rust/crates/core/Cargo.toml

@@ -9,3 +9,7 @@ repository.workspace = true
 serde.workspace = true
 serde_json.workspace = true
 thiserror.workspace = true
+
+[build-dependencies]
+serde.workspace = true
+serde_json.workspace = true