senno 0.2.0

Wits for your backend: provider-agnostic LLM client and streaming tool-calling agent engine, Vertex AI first (Gemini + Claude on Vertex), with built-in history compaction.
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

use crate::error::Error;
use crate::provider::{LlmProvider, LlmStream};
use crate::types::{GenerateRequest, GenerateResponse};
use std::pin::Pin;

use super::config::ResolvedAuth;
use super::{anthropic, gemini};

#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum VertexProvider {
    Gemini,
    Anthropic,
}

pub struct VertexClient {
    pub(crate) http: reqwest::Client,
    pub(crate) auth: ResolvedAuth,
    pub(crate) provider: VertexProvider,
}

impl VertexClient {
    pub(crate) fn new(http: reqwest::Client, auth: ResolvedAuth, provider: VertexProvider) -> Self {
        Self {
            http,
            auth,
            provider,
        }
    }

    pub(crate) async fn authorize(
        &self,
        req: reqwest::RequestBuilder,
    ) -> Result<reqwest::RequestBuilder, Error> {
        match &self.auth {
            ResolvedAuth::ApiKey { api_key } => Ok(req.header("x-goog-api-key", api_key.as_str())),
            ResolvedAuth::ServiceAccount { token_source, .. } => {
                let bearer = token_source
                    .access_token(&["https://www.googleapis.com/auth/cloud-platform"])
                    .await?;
                Ok(req.header("Authorization", format!("Bearer {bearer}")))
            }
        }
    }

    pub(crate) async fn send(
        &self,
        req: reqwest::RequestBuilder,
    ) -> Result<reqwest::Response, Error> {
        let resp = req
            .send()
            .await
            .map_err(|e| Error::provider("vertex-ai", format!("Request failed: {e}")))?;

        if !resp.status().is_success() {
            let status = resp.status();
            let body = resp.text().await.unwrap_or_default();
            return Err(Error::provider(
                "vertex-ai",
                format!("API error ({status}): {body}"),
            ));
        }

        Ok(resp)
    }
}

impl LlmProvider for VertexClient {
    fn generate<'a>(
        &'a self,
        request: &'a GenerateRequest,
    ) -> Pin<Box<dyn std::future::Future<Output = Result<GenerateResponse, Error>> + Send + 'a>>
    {
        Box::pin(async move {
            match self.provider {
                VertexProvider::Gemini => gemini::generate(self, request).await,
                VertexProvider::Anthropic => anthropic::generate(self, request).await,
            }
        })
    }

    fn stream_generate<'a>(
        &'a self,
        request: &'a GenerateRequest,
    ) -> Pin<Box<dyn std::future::Future<Output = Result<LlmStream, Error>> + Send + 'a>> {
        Box::pin(async move {
            match self.provider {
                VertexProvider::Gemini => gemini::stream_generate(self, request).await,
                VertexProvider::Anthropic => anthropic::stream_generate(self, request).await,
            }
        })
    }
}