hermes-rs Agent工作流

Installation

```bash

Build from source

cargo build --release

Or install the CLI crate directly

cargo install --path crates/hermes-cli ```

Tagged releases publish per-platform binaries automatically in the repository's GitHub Releases tab.

Quick Start

```bash

Library Usage

use hermes_core::{
    agent::{HermesAgent, AgentConfig},
    client::{OpenAIClient, ClientConfig},
    tools::{HermesTool, ToolRegistry, ToolContext},
    schema::ToolSchema,
};
use async_trait::async_trait;
use serde_json::Value;

// Define a custom tool
struct MyTool;

#[async_trait]
impl HermesTool for MyTool {
    fn name(&self) -> &str { "my_tool" }
    fn description(&self) -> &str { "My custom tool" }
    fn schema(&self) -> ToolSchema { /* ... */ }

    async fn execute(&self, args: Value, context: ToolContext) -> ToolResult {
        // Your tool logic here
    }
}

// Create the agent
let client = OpenAIClient::new(ClientConfig::default());
let registry = ToolRegistry::new(std::time::Duration::from_secs(30));
registry.register(MyTool).await.unwrap();

let agent = HermesAgent::new(
    AgentConfig::default(),
    client,
    registry,
);

// Run the agent
let response = agent.run("Hello!").await?;
println!("{}", response.content);

Screenshots

Prompt-first landing screen:

Workspace session with conversation, reasoning, and activity panes:

Create a local auth profile that references an environment variable

hermes auth set-api-key openai --env OPENAI_API_KEY

Show provider names, documented auth methods, and Hermes-supported env sources

hermes auth providers

Reference an existing OAuth/ADC bearer token without storing it in config

hermes auth set-bearer-token Google --env GOOGLE_OAUTH_ACCESS_TOKEN --base-url https://generativelanguage.googleapis.com/v1beta ```

Configuration

Hermes reads configuration in this order:

1. --config <path>
2. ./hermes.toml
3. ./.hermes.toml
4. OS config directory (for example ~/.config/hermes/config.toml on Linux)
5. Environment variables
6. CLI flags

Start from the checked-in example file:

cp hermes.example.toml hermes.toml

Configuration is TOML, not YAML. Example:

```toml [client] base_url = "https://api.openai.com/v1" timeout_secs = 60

Optional provider/model rates for spend estimates.

input_cost_per_million = 0.0 output_cost_per_million = 0.0

Or use environment variables:

See hermes.example.toml for the full schema, including MCP, Skills, gateway, and tool/runtime defaults.

CLI Options

hermes [OPTIONS] <COMMAND>

Commands:
  autonomous  Run the autonomous coding loop
  run     Run the agent with a query
  tools   List available tools
  chat    Interactive chat mode
  test    Test a specific tool
  help    Print this message or the help of the given subcommand(s)

Options:
  -v, --verbose           Enable verbose output
  -l, --log-level <LOG>  Log level (debug, info, warn, error) [default: info]
  -c, --config <FILE>    Configuration file path
  --api-key <KEY>        OpenAI API key
  --base-url <URL>       OpenAI base URL
  -m, --model <MODEL>    Model to use [default: gpt-4]
  -i, --max-iterations <N>  Maximum iterations [default: 20]
  --tool-timeout <SECS>  Tool timeout in seconds [default: 30]
  --request-timeout <SECS>  Request timeout in seconds
  --context-window <TOKENS> Context window size
  --max-healing-attempts <N> Maximum self-healing retries
  --stream / --no-stream  Force streaming on or off

Set your API key

export OPENAI_API_KEY=your_api_key_here # PowerShell: $env:OPENAI_API_KEY="..."

api_key = "set me or use OPENAI_API_KEY"

Disposable Repo Workflow

Use a throwaway repository first to validate your autonomous setup end to end:

mkdir hermes-autonomous-sample
cd hermes-autonomous-sample
git init
git checkout -b agent-dev
cp ../hermes-rs/hermes.example.toml ./hermes.toml

Create a minimal TODO.md:

```md