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(agent-harness): update TinyAgents dependency and checkpoint ownership notes
Steven Enamakel99ce3cc10237
diff --git a/crates/openhuman-core/src/agent/subagent_host/mod.rs b/crates/openhuman-core/src/agent/subagent_host/mod.rsindex 454daa636a..7dcbf8b2e1 100644--- a/crates/openhuman-core/src/agent/subagent_host/mod.rs+++ b/crates/openhuman-core/src/agent/subagent_host/mod.rs@@ -33,7 +33,7 @@//! | File | Contents |//! | ----------------- | ----------------------------------------------------------- |//! | `types.rs` | `SubagentRun{Options,Outcome,Error}`, `SubagentMode` |-//! | `ops/` | typed/fork mode and TinyAgents graph route |+//! | `ops/` | typed/fork mode and orchestration route |//! | `tool_prep.rs` | Tool filtering + prompt loading + prompt protocol block |mod autonomous;diff --git a/crates/openhuman-core/src/flows/ops/langfuse_export_tests.rs b/crates/openhuman-core/src/flows/ops/langfuse_export_tests.rsindex e470477e68..8b2c8e78ac 100644--- a/crates/openhuman-core/src/flows/ops/langfuse_export_tests.rs+++ b/crates/openhuman-core/src/flows/ops/langfuse_export_tests.rs@@ -8,8 +8,8 @@ use tinyflows_bus::runs::FlowRunTrigger;/// Builds a minimal observation stream for one node under run/thread ids/// shaped like a real `flows_run` (`thread_id = flow:{id}:{uuid}`).///-/// Built from the serialized TinyAgents graph-event contract consumed by the-/// host's PollRun adapter; no TinyFlows engine types cross this boundary.+/// Built from TinyFlows graph observations consumed by the host's PollRun+/// adapter; these engine types are mapped into TinyAgents' generic span export.fn sample_observations(thread_id: &str) -> Vec<GraphObservation> {let node = NodeId::new("fetch");let mk = |offset: u64, step: usize, ts_ms: u64, event: GraphEvent| GraphObservation {diff --git a/docs/gitbooks/en/developing/architecture/agent-harness.md b/docs/gitbooks/en/developing/architecture/agent-harness.mdindex cae28f9f2a..a91b17f59a 100644--- a/docs/gitbooks/en/developing/architecture/agent-harness.md+++ b/docs/gitbooks/en/developing/architecture/agent-harness.md@@ -105,13 +105,13 @@ Multi-agent orchestration runs on the tinyagents graph layer, through `tinyagent## TinyAgents crate: features and compatibility-OpenHuman depends on the split TinyAgents 2.1 crate family by path into the vendored git submodule `vendor/tinyagents`, `tinyagents-harness` (features `sqlite`, `multimodal`, `builtin-tools`), `tinyagents-definition`, `tinyagents-graph` (`sqlite`), `tinyagents-registry`, `tinyagents-session`, `tinyagents-runtime`, `tinyagents-orchestration` and the optional `tinyagents-live`, plus `tinyinference` from `vendor/tinyagents/vendor/tinyinference`, so SDK changes can be tested in-tree before they go upstream (see [`crates/openhuman-core/Cargo.toml`](https://github.com/tinyhumansai/openhuman/blob/main/crates/openhuman-core/Cargo.toml)). The reasons, so future upgrades do not regress it:+OpenHuman depends on the split TinyAgents crate family by path into the vendored git submodule `vendor/tinyagents`, including `tinyagents-harness` (features `multimodal`, `builtin-tools`), `tinyagents-definition`, `tinyagents-registry`, `tinyagents-session`, `tinyagents-runtime`, `tinyagents-orchestration` and the optional `tinyagents-live`, plus `tinyinference` from `vendor/tinyagents/vendor/tinyinference`, so SDK changes can be tested in-tree before they go upstream (see [`crates/openhuman-core/Cargo.toml`](https://github.com/tinyhumansai/openhuman/blob/main/crates/openhuman-core/Cargo.toml)). TinyFlows owns workflow graph checkpoints and their `graph.*` storage schema. The reasons, so future upgrades do not regress it:- Native TinyAgents model interface, OpenHuman-owned product policy. Every live route is an `Arc<dyn ChatModel<()>>`: TinyAgents OpenAI-compatible clients cover wire-equivalent managed, local, and BYOK routes, while host `ChatModel` implementations cover Claude SDK/Code and Codex-specific transports. OpenHuman still owns credential resolution, OAuth, access gates, endpoint selection, egress disclosure, billing metadata, and error classification.-- `sqlite` feature enabled with one native sqlite chain. OpenHuman's root and Tauri Cargo worlds pin `rusqlite = "=0.40.2"` (`bundled`); the pinned Rust 1.99 toolchain (1.96 or newer) provides the `cfg_select!` macro its build script needs. Both worlds resolve to a single `libsqlite3-sys` chain. Durable graph checkpoints run through TinyAgents' own `SqliteCheckpointer` (see [`orchestration/delegation.rs`](https://github.com/tinyhumansai/openhuman/blob/main/crates/openhuman-core/src/agent/orchestration/delegation.rs)).+- One native SQLite chain serves OpenHuman persistence. OpenHuman's root and Tauri Cargo worlds pin `rusqlite = "=0.40.2"` (`bundled`); the pinned Rust 1.99 toolchain (1.96 or newer) provides the `cfg_select!` macro its build script needs. Both worlds resolve to a single `libsqlite3-sys` chain. TinyFlows persists workflow checkpoints through its driver-owned checkpointer and schema.- WhatsApp Web lives in the TinyChannels module (`tinychannels/whatsapp-web`), not in OpenHuman. OpenHuman has no `whatsapp-web` feature and no `whatsapp-rust` or `wacore` dependency in its normal graph. The module's storage bridge is documented in the tinychannels repository.- No scripting runtime. There is no `repl` or Rhai scripting feature and no `rhai_workflows` tool. Rhai still appears as a dependency of `tinyflows` for its own JSON-query stack, unrelated to the agent harness.-- Ownership map: model construction → `inference::provider::create_chat_model*`; durable graph checkpoints → TinyAgents' `SqliteCheckpointer`; generic detached executor state → `DetachedTaskRegistry`; controller-facing durability → OpenHuman SQL/JSON run ledgers (`running_subagents`, `workflow_runs`, `agent_teams`, `command_center`). The generic harness/graph/middleware/event primitives are used as-is.+- Ownership map: model construction → `inference::provider::create_chat_model*`; workflow checkpoints → TinyFlows' driver schema; generic detached executor state → `DetachedTaskRegistry`; controller-facing durability → OpenHuman SQL/JSON run ledgers (`running_subagents`, `workflow_runs`, `agent_teams`, `command_center`). The generic harness/middleware/event primitives are used as-is.The agent harness is the runtime that turns a user message (or a webhook fire, or a cron tick) into a complete, tool-using LLM interaction. It owns the tool-call loop, sub-agent dispatch, the trigger-triage pipeline, and the hook surface around them. It does not own provider HTTP transport, tool implementations, prompt-section assembly or memory storage. Those are separate domains the harness composes.