hematite-cli 0.7.1

Senior SysAdmin, Network Admin, and Software Engineer living in your terminal. A high-precision local AI agent harness for LM Studio, Ollama, and other local OpenAI-compatible runtimes that runs 100% on your own silicon. Reads repos, edits files, runs builds, and inspects the machine it is running on—including full network state and workstation telemetry.
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

/// Staged triage engine for /diagnose.
///
/// Phase 1 (harness): runs health_report deterministically and identifies
/// which follow-up topics are warranted based on the findings.
/// Phase 2 (agent): receives a pre-built instruction naming the exact topics
/// to investigate, so the model synthesizes rather than orchestrates.

/// Parse health_report text and return the follow-up inspect_host topics
/// that the harness determined are worth drilling into. Skips dev-environment
/// "not installed" warnings (those aren't system health issues).
pub fn triage_follow_up_topics(health_output: &str) -> Vec<&'static str> {
    let lower = health_output.to_ascii_lowercase();
    let mut topics: Vec<&'static str> = Vec::new();

    let action_required = lower.contains("action required");
    let worth_a_look = lower.contains("worth a look");
    if !action_required && !worth_a_look {
        return topics; // ALL GOOD — no follow-up needed
    }

    // Disk / storage — flag when disk appears in needs_fix or watch tier
    if lower.contains("[!]") && (lower.contains("disk") || lower.contains("drive")) {
        topics.push("storage");
        topics.push("disk_health");
    } else if lower.contains("[-]") && (lower.contains("disk") || lower.contains("drive")) {
        topics.push("storage");
    }

    // RAM / memory pressure
    if (lower.contains("[!]") || lower.contains("[-]")) && lower.contains("ram") {
        topics.push("resource_load");
        topics.push("processes");
    }

    // Windows event log errors — always worth drilling into
    if lower.contains("critical") || lower.contains("error event") {
        if lower.contains("event") {
            topics.push("log_check");
        }
    }

    // Critical services stopped
    if lower.contains("[!]") && lower.contains("service") {
        topics.push("services");
    } else if lower.contains("[-]") && lower.contains("service") {
        topics.push("services");
    }

    // Security / Defender issues
    if (lower.contains("[!]") || lower.contains("[-]"))
        && (lower.contains("defender") || lower.contains("firewall") || lower.contains("security"))
    {
        topics.push("security");
    }

    // Network unreachable or high latency
    if lower.contains("[!]") && lower.contains("internet connectivity") {
        topics.push("connectivity");
        topics.push("network");
    } else if lower.contains("[-]") && lower.contains("internet connectivity") {
        topics.push("connectivity");
    }

    // Pending reboot
    if lower.contains("pending reboot") {
        topics.push("pending_reboot");
    }

    // Thermal elevated or very high
    if (lower.contains("[!]") || lower.contains("[-]"))
        && (lower.contains("thermal") || lower.contains("°c"))
    {
        topics.push("thermal");
        topics.push("overclocker");
    }

    // Deduplicate while preserving order
    let mut seen = std::collections::HashSet::new();
    topics.retain(|t| seen.insert(*t));

    topics
}

/// Build the agent instruction for phase 2 of /diagnose.
/// The harness has already run health_report; this tells the agent exactly
/// which topics to investigate and how to synthesize the results.
pub fn build_diagnose_instruction(health_output: &str, follow_up_topics: &[&str]) -> String {
    if follow_up_topics.is_empty() {
        return format!(
            "DIAGNOSE MODE — triage complete.\n\n\
             Health report:\n{}\n\n\
             The machine is in good health. Summarize the key findings for the user \
             in 2-3 sentences and confirm no action is needed.",
            health_output
        );
    }

    let topic_list = follow_up_topics
        .iter()
        .enumerate()
        .map(|(i, t)| format!("{}. inspect_host(topic=\"{}\")", i + 1, t))
        .collect::<Vec<_>>()
        .join("\n");

    format!(
        "DIAGNOSE MODE — harness triage identified {} area(s) to investigate.\n\n\
         Health report (already run by harness):\n{}\n\n\
         PROTOCOL — follow this exactly:\n\
         Call each topic below in order:\n{}\n\n\
         After all calls complete:\n\
         - Write a numbered fix plan grounded in the tool output\n\
         - Lead with the most critical issue first\n\
         - Every step must reference specific data from the results (exact path, count, service name, etc.)\n\
         - No generic advice — only steps that address what the tools actually found\n\
         - If a finding needs a restart or elevated privileges, say so explicitly",
        follow_up_topics.len(),
        health_output,
        topic_list
    )
}