mermaid_cli/models/
config.rs

1//! Unified configuration system for models and backends
2//!
3//! Replaces the fragmented app::Config + models::ModelConfig split
4//! with a single, coherent, backend-agnostic configuration structure.
5
6use crate::constants::{DEFAULT_MAX_TOKENS, DEFAULT_TEMPERATURE};
7use crate::prompts;
8use serde::{Deserialize, Serialize};
9use std::collections::HashMap;
10
11/// Unified model configuration
12#[derive(Debug, Clone, Serialize, Deserialize)]
13pub struct ModelConfig {
14    /// Model identifier (provider/model or just model name)
15    /// Examples: "ollama/qwen3-coder:30b", "qwen3-coder:30b", "gpt-4"
16    pub model: String,
17
18    /// Temperature (0.0-2.0, controls randomness)
19    #[serde(default = "default_temperature")]
20    pub temperature: f32,
21
22    /// Maximum tokens to generate
23    #[serde(default = "default_max_tokens")]
24    pub max_tokens: usize,
25
26    /// System prompt override (None = use default)
27    pub system_prompt: Option<String>,
28
29    /// Enable thinking mode for models that support it (e.g., kimi, qwen3)
30    /// Some(true) = explicitly enabled, Some(false) = explicitly disabled, None = model default
31    #[serde(default = "default_thinking_enabled")]
32    pub thinking_enabled: Option<bool>,
33
34    /// Whether this is a subagent context (excludes the agent tool to prevent nesting).
35    /// Runtime-only flag -- never persisted to disk.
36    #[serde(skip)]
37    pub is_subagent: bool,
38
39    /// Backend-specific options (provider name -> key/value pairs)
40    /// Example: {"ollama": {"num_gpu": "10", "num_ctx": "8192"}}
41    #[serde(default)]
42    pub backend_options: HashMap<String, HashMap<String, String>>,
43}
44
45impl Default for ModelConfig {
46    fn default() -> Self {
47        Self {
48            model: "ollama/tinyllama".to_string(),
49            temperature: default_temperature(),
50            max_tokens: default_max_tokens(),
51            system_prompt: Some(prompts::get_system_prompt()),
52            thinking_enabled: Some(true),
53            is_subagent: false,
54            backend_options: HashMap::new(),
55        }
56    }
57}
58
59impl ModelConfig {
60    /// Get a backend-specific option
61    pub fn get_backend_option(&self, backend: &str, key: &str) -> Option<&String> {
62        self.backend_options.get(backend)?.get(key)
63    }
64
65    /// Get backend option as integer
66    pub fn get_backend_option_i32(&self, backend: &str, key: &str) -> Option<i32> {
67        self.get_backend_option(backend, key)?.parse::<i32>().ok()
68    }
69
70    /// Get backend option as boolean
71    pub fn get_backend_option_bool(&self, backend: &str, key: &str) -> Option<bool> {
72        self.get_backend_option(backend, key)?.parse::<bool>().ok()
73    }
74
75    /// Set a backend-specific option
76    pub fn set_backend_option(&mut self, backend: String, key: String, value: String) {
77        self.backend_options
78            .entry(backend)
79            .or_default()
80            .insert(key, value);
81    }
82
83    /// Extract Ollama-specific options
84    pub fn ollama_options(&self) -> OllamaOptions {
85        OllamaOptions {
86            num_gpu: self.get_backend_option_i32("ollama", "num_gpu"),
87            num_thread: self.get_backend_option_i32("ollama", "num_thread"),
88            num_ctx: self.get_backend_option_i32("ollama", "num_ctx"),
89            numa: self.get_backend_option_bool("ollama", "numa"),
90        }
91    }
92}
93
94/// Ollama-specific options (extracted from backend_options)
95#[derive(Debug, Clone, Default)]
96pub struct OllamaOptions {
97    pub num_gpu: Option<i32>,
98    pub num_thread: Option<i32>,
99    pub num_ctx: Option<i32>,
100    pub numa: Option<bool>,
101}
102
103/// Backend connection configuration
104#[derive(Debug, Clone, Serialize, Deserialize)]
105pub struct BackendConfig {
106    /// Ollama server URL (default: http://localhost:11434)
107    #[serde(default = "default_ollama_url")]
108    pub ollama_url: String,
109
110    /// Connection timeout in seconds
111    #[serde(default = "default_timeout")]
112    pub timeout_secs: u64,
113
114    /// Max idle connections per host
115    #[serde(default = "default_max_idle")]
116    pub max_idle_per_host: usize,
117}
118
119impl Default for BackendConfig {
120    fn default() -> Self {
121        Self {
122            ollama_url: default_ollama_url(),
123            timeout_secs: default_timeout(),
124            max_idle_per_host: default_max_idle(),
125        }
126    }
127}
128
129// Default value functions
130fn default_temperature() -> f32 {
131    DEFAULT_TEMPERATURE
132}
133
134fn default_max_tokens() -> usize {
135    DEFAULT_MAX_TOKENS
136}
137
138fn default_ollama_url() -> String {
139    std::env::var("OLLAMA_HOST").unwrap_or_else(|_| "http://localhost:11434".to_string())
140}
141
142fn default_timeout() -> u64 {
143    10
144}
145
146fn default_max_idle() -> usize {
147    10
148}
149
150fn default_thinking_enabled() -> Option<bool> {
151    Some(true)
152}
mermaid_cli/models/config.rs

mermaid_cli/models/
config.rs
