Pre-launch — Gluecron is in final validation. Public signups and git hosting for non-owner users open after launch review.
CodeIssuesDiscussionsWikiPull RequestsProjectsCommitsActionsReleasesContributorsPulse● GatesSecuritySettingsDeploymentsPipelineInsightsAgents✨ Explain✨ Ask AI✨ Workspace✨ Spec✨ Tests▓ Debt Map✨ NL Search🏛 Archaeology
claude/adoring-hopper-5x74bqclaude/affectionate-feynman-ykrf1hclaude/architecture-audit-design-wxprenclaude/build-status-update-3MXsfclaude/charming-meitner-mllb5rclaude/compare-gate-gluecron-s4mFQclaude/confident-faraday-tikcwbclaude/continue-work-XMTlIclaude/crontech-gluecron-deploy-7MIECclaude/crontech-platform-setup-SeKfwclaude/design-2026claude/ecstatic-ptolemy-jMdigclaude/enhance-github-integration-QNHdGclaude/fix-aa-loop-issue-PonMQclaude/fix-actions-and-processclaude/fix-desktop-errors-XqoW8claude/fix-red-workflowsclaude/fix-website-access-6FKJNclaude/gatetest-integration-hardeningclaude/github-audit-improvements-bDFr9claude/gluecron-launch-status-FoMRlclaude/hopeful-lamport-olfCTclaude/issue-to-pr-and-protectionsclaude/jolly-heisenberg-2sg1Qclaude/launch-preparation-QmTb6claude/new-session-xk1l7claude/plan-platform-architecture-kkN4yclaude/platform-analysis-roadmap-1nUGLclaude/platform-launch-assessment-8dWV8claude/polish-platform-release-AeDrUclaude/resume-previous-work-KzyLwclaude/review-crontech-handoff-qYEVqclaude/review-project-completeness-lHhS2claude/review-readme-docs-ulqPKclaude/serene-edison-rj87weclaude/setup-multi-repo-dev-BCwNQclaude/ship-fixes-and-tests-Jvz1cclaude/site-audit-competitive-pctlwgclaude/site-migration-vercel-XstpKclaude/standalone-product-repos-XHFTDcopilot/feat-smart-empty-states-keyboard-first-enhancementcopilot/feat-smart-morning-digest-review-context-restorecopilot/fix-and-process-workflowscopilot/update-ai-powered-code-reviewfeat/debt-mapfeat/push-policy-codeowners-hardeningfeat/smart-digest-contextfeat/stage-impactfeat/t1-secret-migrationfeat/u-polishfeat/w-self-hostfeat/w2-claude-configfix/agent-journey-orphan-sweepgatetest/auto-fix-1776586424172gatetest/auto-fix-1776586534814gatetest/auto-fix-1776590685143gatetest/auto-fix-1776590808199mainops/redeploy-retriggerstyle/dxt-cta-themeworktree-agent-a3377aad30d55da26worktree-agent-a7ef607b7ee1d6c74
ai-review.ts3.0 KB · 125 lines
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
/**
 * AI-powered code review using Claude.
 *
 * Generates inline review comments on pull request diffs.
 * Reviews are posted as PR comments with isAiReview=true.
 */

import Anthropic from "@anthropic-ai/sdk";
import { config } from "./config";

interface ReviewComment {
  filePath: string;
  lineNumber: number | null;
  body: string;
}

interface ReviewResult {
  summary: string;
  comments: ReviewComment[];
  approved: boolean;
}

let _client: Anthropic | null = null;

function getClient(): Anthropic {
  if (!_client) {
    if (!config.anthropicApiKey) {
      throw new Error("ANTHROPIC_API_KEY is not set");
    }
    _client = new Anthropic({ apiKey: config.anthropicApiKey });
  }
  return _client;
}

/**
 * Run AI code review on a PR diff.
 */
export async function reviewDiff(
  repoFullName: string,
  prTitle: string,
  prBody: string | null,
  baseBranch: string,
  headBranch: string,
  diffText: string
): Promise<ReviewResult> {
  const client = getClient();

  const message = await client.messages.create({
    model: "claude-sonnet-4-20250514",
    max_tokens: 4096,
    messages: [
      {
        role: "user",
        content: `You are reviewing a pull request on the repository "${repoFullName}".

**PR Title:** ${prTitle}
**PR Description:** ${prBody || "(none)"}
**Base branch:** ${baseBranch}
**Head branch:** ${headBranch}

Review the following diff. Look for:
- Bugs, logic errors, or potential runtime failures
- Security vulnerabilities (injection, XSS, auth bypasses, secrets in code)
- Performance issues (N+1 queries, unnecessary allocations, blocking I/O)
- Missing error handling at system boundaries
- Breaking changes or API contract violations

Do NOT comment on style, formatting, naming, missing docs, or minor nitpicks. Only flag issues that could cause real problems.

Respond in JSON format:
{
  "summary": "1-3 sentence overall assessment",
  "approved": true/false,
  "comments": [
    {
      "filePath": "path/to/file.ts",
      "lineNumber": 42,
      "body": "Explain the issue and suggest a fix"
    }
  ]
}

If the diff looks clean, return approved: true with an empty comments array.

\`\`\`diff
${diffText.slice(0, 100000)}
\`\`\``,
      },
    ],
  });

  const text =
    message.content[0].type === "text" ? message.content[0].text : "";

  try {
    // Extract JSON from response (may be wrapped in markdown code block)
    const jsonMatch = text.match(/\{[\s\S]*\}/);
    if (!jsonMatch) {
      return {
        summary: "AI review completed but could not parse structured output.",
        comments: [],
        approved: true,
      };
    }
    const parsed = JSON.parse(jsonMatch[0]);
    return {
      summary: parsed.summary || "Review complete.",
      comments: Array.isArray(parsed.comments) ? parsed.comments : [],
      approved: parsed.approved !== false,
    };
  } catch {
    return {
      summary: text.slice(0, 500),
      comments: [],
      approved: true,
    };
  }
}

/**
 * Check if AI review is available (API key configured).
 */
export function isAiReviewEnabled(): boolean {
  return !!config.anthropicApiKey;
}