kumo 0.2.5

An async web crawling framework for Rust — Scrapy for Rust
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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137

use super::{TokenUsage, UsageCounters, shared};
use crate::error::KumoError;
use async_trait::async_trait;
use rig::client::{CompletionClient, Nothing};
use rig::completion::CompletionModel;
use rig::providers::ollama;
use serde_json::Value;
use std::sync::Arc;

/// LLM client for local Ollama, powered by rig-core.
///
/// Uses `OLLAMA_API_BASE_URL` env var (default: `http://localhost:11434`).
/// For a custom URL, use [`OllamaClient::with_url`].
///
/// # Example
/// ```rust,ignore
/// let client = OllamaClient::new().model("llama3.2");
///
/// let items: Vec<Item> = res.extract(&client).await?;
/// ```
pub struct OllamaClient {
    inner: ollama::Client,
    model: String,
    system_prompt: Option<String>,
    prompt_template: Option<String>,
    strip_scripts: bool,
    max_tokens: u64,
    usage: Arc<UsageCounters>,
}

impl OllamaClient {
    /// Create a client using the default Ollama URL (`http://localhost:11434`).
    pub fn new() -> Self {
        let inner = ollama::Client::builder()
            .api_key(Nothing)
            .build()
            .expect("failed to build Ollama client");
        Self {
            inner,
            model: "llama3.2".into(),
            system_prompt: None,
            prompt_template: None,
            strip_scripts: false,
            max_tokens: 4096,
            usage: UsageCounters::new(),
        }
    }

    /// Create a client pointing at a custom Ollama base URL.
    pub fn with_url(base_url: impl AsRef<str>) -> Self {
        let inner = ollama::Client::builder()
            .api_key(Nothing)
            .base_url(base_url.as_ref())
            .build()
            .expect("failed to build Ollama client");
        Self {
            inner,
            model: "llama3.2".into(),
            system_prompt: None,
            prompt_template: None,
            strip_scripts: false,
            max_tokens: 4096,
            usage: UsageCounters::new(),
        }
    }

    pub fn model(mut self, model: impl Into<String>) -> Self {
        self.model = model.into();
        self
    }

    pub fn system_prompt(mut self, prompt: impl Into<String>) -> Self {
        self.system_prompt = Some(prompt.into());
        self
    }

    pub fn prompt_template(mut self, template: impl Into<String>) -> Self {
        self.prompt_template = Some(template.into());
        self
    }

    pub fn strip_scripts_and_styles(mut self, yes: bool) -> Self {
        self.strip_scripts = yes;
        self
    }

    pub fn max_tokens(mut self, n: u64) -> Self {
        self.max_tokens = n;
        self
    }

    /// Returns the cumulative token usage across all `extract` calls on this client.
    pub fn total_usage(&self) -> TokenUsage {
        self.usage.snapshot()
    }
}

impl Default for OllamaClient {
    fn default() -> Self {
        Self::new()
    }
}

#[async_trait]
impl super::LlmClient for OllamaClient {
    async fn extract_json(
        &self,
        schema: &Value,
        html: &str,
    ) -> Result<(Value, TokenUsage), KumoError> {
        let args = shared::build_extract_args(
            schema,
            html,
            self.system_prompt.as_deref(),
            self.prompt_template.as_deref(),
            self.strip_scripts,
        );

        let model = self.inner.completion_model(&self.model);
        let request = model
            .completion_request(args.user_content)
            .preamble(args.system)
            .tool(args.tool)
            .max_tokens(self.max_tokens)
            .build();

        let resp = model
            .completion(request)
            .await
            .map_err(|e| shared::llm_err(format!("Ollama error — {e}")))?;

        let usage = TokenUsage::from_rig(&resp.usage);
        let value = shared::extract_tool_input(resp.choice, "extract")?;
        self.usage.add(&usage);
        Ok((value, usage))
    }
}