rho-coding-agent 1.16.0

A lightweight agent harness inspired by Pi
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

//! Background polling for model metadata and update notices.

use futures_util::FutureExt;
use rho_providers::model::models_dev::fetch_model_metadata;
use rho_providers::model::ReasoningRequestSource::PersistedOrDefault;

use crate::credential_store::build_provider;

use super::{reasoning_metadata, App, Entry, InteractiveRuntime};

impl App {
    pub(super) fn poll_update_notice(&mut self) {
        let Some(handle) = self.pending_update_notice.as_mut() else {
            return;
        };
        let Some(result) = handle.now_or_never() else {
            return;
        };
        self.pending_update_notice = None;
        if let Ok(Some(notice)) = result {
            self.info.services.update_notice = Some(notice);
        }
    }

    /// Returns whether the runtime accepted the context window.
    fn apply_context_window(
        &mut self,
        agent: &mut InteractiveRuntime,
        context_window: Option<u64>,
    ) -> bool {
        if let Err(err) = agent.set_context_window(context_window) {
            self.insert_entry(&Entry::Error(format!(
                "could not apply the model context window: {err}"
            )));
            return false;
        }
        true
    }

    pub(super) fn start_model_metadata_fetch(&mut self, agent: &mut InteractiveRuntime) {
        if let Some(handle) = self.pending_model_metadata.take() {
            handle.abort();
        }
        self.pending_model_metadata_reasoning = None;
        if let Some((metadata, metadata_is_current)) = reasoning_metadata::cached_metadata(
            &self.info.runtime.provider,
            &self.info.runtime.model,
        ) {
            if self.apply_context_window(agent, metadata.display_context_window()) {
                let reasoning_metadata_complete = metadata.reasoning_metadata_complete;
                self.model_metadata = Some(metadata);
                if reasoning_metadata_complete && metadata_is_current {
                    return;
                }
            }
            // Failed apply: leave any prior cache alone and fall through to fetch.
        } else {
            let _ = self.apply_context_window(agent, None);
            self.model_metadata = None;
        }
        let provider = self.info.runtime.provider.clone();
        let model = self.info.runtime.model.clone();
        self.pending_model_metadata_reasoning = Some((
            self.info.runtime.reasoning,
            self.info.runtime.reasoning_source,
        ));
        self.pending_model_metadata = Some(tokio::spawn(async move {
            fetch_model_metadata(&provider, &model).await
        }));
    }

    pub(super) fn poll_model_metadata_fetch(&mut self, agent: &mut InteractiveRuntime) {
        let Some(handle) = self.pending_model_metadata.as_mut() else {
            return;
        };
        if !handle.is_finished() {
            return;
        }
        if let Some(handle) = self.pending_model_metadata.take() {
            let reasoning_at_fetch_start = self.pending_model_metadata_reasoning.take();
            if let Some(Ok(Some(metadata))) = handle.now_or_never() {
                if !self.apply_context_window(agent, metadata.display_context_window()) {
                    // Keep prior metadata until a later fetch can apply cleanly.
                    return;
                }
                let capabilities = metadata.reasoning_capabilities();
                let resolved = reasoning_metadata::resolve_fetched_reasoning(
                    &capabilities,
                    self.info.runtime.reasoning,
                    reasoning_at_fetch_start,
                );
                let reasoning = resolved.effective;
                if let Some(requested) = resolved.rejected {
                    self.insert_entry(&Entry::Error(format!(
                        "reasoning level '{requested}' is not supported by {}/{}; restored '{reasoning}'",
                        self.info.runtime.provider, self.info.runtime.model
                    )));
                }
                let provider_updated = match build_provider(
                    &self.info.runtime.provider,
                    &self.info.runtime.model,
                    reasoning,
                ) {
                    Ok(provider) => match agent.replace_provider(provider, reasoning) {
                        Ok(_) => true,
                        Err(err) => {
                            self.insert_entry(&Entry::Error(format!(
                                "could not apply model reasoning metadata: {err}"
                            )));
                            false
                        }
                    },
                    Err(err) => {
                        self.insert_entry(&Entry::Error(format!(
                            "could not apply model reasoning metadata: {err}"
                        )));
                        false
                    }
                };
                if provider_updated && reasoning != self.info.runtime.reasoning {
                    self.info.set_reasoning(reasoning, PersistedOrDefault);
                    if let Err(err) = self.info.services.config_repository.update(|config| {
                        config.reasoning = reasoning;
                    }) {
                        self.insert_entry(&Entry::Error(format!(
                            "could not save normalized reasoning: {err}"
                        )));
                    }
                }
                self.model_metadata = Some(metadata);
            }
        }
    }
}