The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.
Rust agent-orchestrationai-agentsai-assistantdesktopllmlocal-firstmcppersonal-aiprivacyrustsecond-braintauri
docs: update embedding host and egress docs
Steven Enamakel3cee4fa2c8d9
diff --git a/crates/openhuman-core/src/config/schema/load/mod.rs b/crates/openhuman-core/src/config/schema/load/mod.rsindex 2af2015d90..798d102983 100644--- a/crates/openhuman-core/src/config/schema/load/mod.rs+++ b/crates/openhuman-core/src/config/schema/load/mod.rs@@ -45,10 +45,10 @@ pub(crate) use dirs::default_root_dir_name_pub as default_root_dir_name;// re-export; only the load_tests module needs it visible at this level.#[cfg(test)]pub(crate) use dirs::read_active_user_id_checked;-// Non-test: the keyless cloud embedder (`embeddings::cloud_adapter`)-// resolves its `OPENHUMAN_WORKSPACE` credential scope through the same-// workspace→config-dir mapping `config::load` uses, so a legacy `.../workspace`-// override lands on the sibling `.openhuman` root that holds `auth-profiles.json`.+// Non-test: the managed embedding factory resolves its credential scope+// through the same workspace→config-dir mapping `config::load` uses, so a+// legacy `.../workspace` override lands on the sibling `.openhuman` root that+// holds `auth-profiles.json`.pub(crate) use dirs::resolve_config_dir_for_workspace;#[cfg(test)]pub(crate) use dirs::{diff --git a/crates/openhuman-core/src/inference/embedding_host/README.md b/crates/openhuman-core/src/inference/embedding_host/README.mdindex facbf824a3..d203b036d9 100644--- a/crates/openhuman-core/src/inference/embedding_host/README.md+++ b/crates/openhuman-core/src/inference/embedding_host/README.md@@ -28,16 +28,17 @@ Declared in [`mod.rs`](./mod.rs) and resolved by [`factory.rs`](./factory.rs):| --- | --- || `mod.rs` | Module docs and re-exports. Declares the provider list above. || `factory.rs` | OpenHuman's binding onto TinyInference's embedding factory: `create_embedding_provider_with_credentials`, `create_embedding_provider_with_config`, and `default_embedding_provider_with_config` (falls back to a `Noop` model if the configured provider fails to build). |-| [`cloud_adapter.rs`](./cloud_adapter.rs) | `OpenHumanCloudEmbeddingModel`: the managed-provider transport, with OpenHuman's own bearer-token and egress-guard wiring around the crate-owned `CloudEmbeddingModel`. |+| `factory.rs` | The managed-model closure keeps OpenHuman's registered provider, config-scoped credentials and approval context at the host boundary while using TinyInference's unavailable placeholder when no provider is registered. || [`rpc.rs`](./rpc.rs) + `rpc/` (`api_keys.rs`, `embed.rs`, `settings.rs`) | The `embeddings` RPC handlers: get/update settings, set/clear API key, embed text, test a connection. || [`schemas.rs`](./schemas.rs) | Controller schemas and thin handlers for the `embeddings` RPC namespace, registered as `all_embeddings_controller_schemas` / `all_embeddings_registered_controllers`. |## The model contract lives in TinyInferenceCallers hold `tinyinference_embeddings::EmbeddingModel` directly (batches are-`&[String]`). The trait, `format_embedding_signature` and `NoopEmbeddingModel`-are owned by `tinyinference-embeddings`; this host only binds credentials,-rate limits and the managed cloud transport onto it.+`&[String]`). The trait, verification helpers, `UnavailableEmbeddingModel`,+`format_embedding_signature` and `NoopEmbeddingModel` are owned by+`tinyinference-embeddings`; this host binds credentials, rate limits and the+registered managed provider onto them.## How it fitsdiff --git a/crates/openhuman-core/src/security/egress/README.md b/crates/openhuman-core/src/security/egress/README.mdindex 854159a021..16f493a7e2 100644--- a/crates/openhuman-core/src/security/egress/README.md+++ b/crates/openhuman-core/src/security/egress/README.md@@ -67,9 +67,9 @@ Both `enforce_egress` and `local_only_tool_block` read the live mode via`integrations/composio/execute_dispatch.rs`: backend and Composio calls.- `tools/impl/network/{curl,http_request,web_fetch}.rs`: agent network-fetchtools.-- `search/tools/{tavily/client,exa}.rs`, `embeddings/cloud_adapter.rs`,- `inference/provider/factory.rs` and its `factory/` submodules: search and- cloud inference/embedding providers.+- `search/tools/{tavily/client,exa}.rs`, `inference/provider/factory.rs` and+ its `factory/` submodules, and the TinyHumans managed embedding provider:+ search and hosted inference/embedding calls.- `web_chat/event_bus.rs`: subscribes to `ExternalTransferPending` andbridges it to the `external_transfer_pending` socket event for the frontend.