everruns-runtime 0.11.0

Public in-process runtime for embedding Everruns harnesses
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93

# everruns-runtime

> Embed Everruns agents directly in your Rust process — no server, worker, or database required.

[![Crates.io](https://img.shields.io/crates/v/everruns-runtime.svg)](https://crates.io/crates/everruns-runtime)
[![Documentation](https://docs.rs/everruns-runtime/badge.svg)](https://docs.rs/everruns-runtime)
[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](https://github.com/everruns/everruns/blob/main/LICENSE)

`everruns-runtime` is the in-process entrypoint for embedding Everruns. It runs a
harness entirely inside your own process — with in-memory stores by default —
using the same `input → reason → act` loop as worker- and server-backed
execution, but without the durable engine, gRPC worker boundary, or
control-plane server. It is the recommended starting point for trying Everruns,
writing tests, and embedding agents in your own application.

Part of the [Everruns](https://everruns.com) ecosystem — the durable agentic
harness engine for building unstoppable agents. The runtime builds on the
contracts in [`everruns-core`](https://crates.io/crates/everruns-core) and pairs
with provider crates such as
[`everruns-openai`](https://crates.io/crates/everruns-openai) and
[`everruns-anthropic`](https://crates.io/crates/everruns-anthropic).

## Quick Example

```rust
use everruns_core::{
    CapabilityRegistry, DriverRegistry, LlmProviderType, ModelWithProvider, PlatformDefinition,
};
use everruns_core::capabilities::TestMathCapability;
use everruns_runtime::InProcessRuntimeBuilder;

# async fn example() -> Result<(), Box<dyn std::error::Error>> {
let mut capabilities = CapabilityRegistry::new();
capabilities.register(TestMathCapability);

let platform = PlatformDefinition::new(capabilities, DriverRegistry::new());
let runtime = InProcessRuntimeBuilder::new()
    .platform_definition(platform)
    .llm_sim(everruns_core::llmsim_driver::LlmSimConfig::fixed("4"))
    .default_model(ModelWithProvider {
        model: "llmsim-model".into(),
        provider_type: LlmProviderType::LlmSim,
        api_key: Some("fake-key".into()),
        base_url: None,
    })
    .single_session(|s| {
        s.harness("math", "You are a calculator.")
            .with_capability("test_math")
            .agent("math-agent", "Use tools when useful.")
            .session_title("Math example")
    })
    .build()
    .await?;

let session_id = runtime.default_session_id().expect("single_session id");
let result = runtime.run_text_turn(session_id, "What is 2 + 2?").await?;

assert!(result.success);
# Ok(())
# }
```

The example above uses the built-in deterministic LLM simulator so it runs with
no API key. Swap in a real provider (for example
[`everruns-openai`](https://crates.io/crates/everruns-openai)) to talk to a
hosted model.

Runnable examples ship with the crate:

```text
cargo run -p everruns-runtime --example in_process_runtime
cargo run -p everruns-runtime --example inspect_context
cargo run -p everruns-runtime --example real_disk_file_system_tools
```

## What It Provides

- `InProcessRuntimeBuilder` for declaring platforms, harnesses, agents, and sessions in code
- In-memory stores for local development and tests, with hooks to plug in custom backends
- Real-disk workspace wiring so file-backed agents can read and write an actual directory
- Turn-context inspection before and after a turn executes
- Host-phase helpers reused by durable and server-backed execution

## Documentation

- [API reference (docs.rs)](https://docs.rs/everruns-runtime)
- [Runtime — embed Everruns in your process](https://docs.everruns.com/features/runtime/)
- [Tutorial: build your first agent](https://docs.everruns.com/tutorials/building-agents-using-sdk/)
- [Everruns documentation](https://docs.everruns.com)

## License

Licensed under the [MIT License](https://github.com/everruns/everruns/blob/main/LICENSE).