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-client.ts3.8 KB · 126 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
126
/**
 * Shared Anthropic client + helpers for all AI-powered features.
 * Centralised here so model choice, caching, and key handling live in one place.
 */

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

let _client: Anthropic | null = null;

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

export function isAiAvailable(): boolean {
  return !!config.anthropicApiKey;
}

/** Primary model for all AI features — code understanding, review, generation */
export const MODEL_SONNET = "claude-sonnet-4-6";
/** Light-task model — never reference directly; route through modelForTask() */
export const MODEL_HAIKU = "claude-haiku-4-5-20251001";

/**
 * Task → model routing.
 *
 * Every AI feature names its task here and asks `modelForTask()` which model
 * to run. Routing policy is deliberately conservative:
 *
 *   - Haiku is allowed ONLY for short, low-stakes outputs that a human
 *     always reviews and can trivially override (commit-message drafts,
 *     issue/PR triage suggestions, label suggestions). Worst-case failure
 *     is a bad suggestion someone ignores.
 *   - EVERYTHING else — anything that writes code, judges code, or produces
 *     user-facing documents — stays on Sonnet. Unknown tasks default to
 *     Sonnet too, so a typo can never silently downgrade quality.
 *
 * Kill-switch: set AI_FORCE_SONNET=1 to route every task to Sonnet. The env
 * var is read at call time, so flipping it takes effect on the next request
 * without a restart.
 */
export type AiTask =
  // Haiku-eligible (short, human-reviewed, low-stakes suggestions)
  | "commit-message"
  | "issue-triage"
  | "pr-triage"
  | "label-suggest"
  // Sonnet-only (writes or judges code, or produces user-facing docs)
  | "code-review"
  | "code-completion"
  | "spec-to-pr"
  | "ci-heal"
  | "pr-summary"
  | "changelog";

/** Tasks allowed to run on Haiku. Additions require explicit owner sign-off. */
const HAIKU_ALLOWLIST: ReadonlySet<AiTask> = new Set<AiTask>([
  "commit-message",
  "issue-triage",
  "pr-triage",
  "label-suggest",
]);

/**
 * Resolve the model for a task. Allowlisted light tasks get Haiku; everything
 * else (including unknown task strings) gets Sonnet. `AI_FORCE_SONNET=1`
 * forces Sonnet for all tasks.
 */
export function modelForTask(task: AiTask): string {
  // Read at call time so the kill-switch works without a restart.
  if (process.env.AI_FORCE_SONNET === "1") return MODEL_SONNET;
  return HAIKU_ALLOWLIST.has(task) ? MODEL_HAIKU : MODEL_SONNET;
}

/**
 * Extract text content from an Anthropic message response.
 */
export function extractText(
  message: Anthropic.Messages.Message
): string {
  for (const block of message.content) {
    if (block.type === "text") return block.text;
  }
  return "";
}

/**
 * Safely parse JSON from a Claude response that may include surrounding prose
 * or a ```json code block.
 */
export function parseJsonResponse<T = unknown>(text: string): T | null {
  // Prefer ```json blocks
  const fenced = text.match(/```(?:json)?\s*(\{[\s\S]*?\}|\[[\s\S]*?\])\s*```/);
  const candidate = fenced ? fenced[1] : null;
  if (candidate) {
    try {
      return JSON.parse(candidate) as T;
    } catch {
      // fall through
    }
  }
  // Fall back to first top-level {...} or [...]
  const braceMatch = text.match(/\{[\s\S]*\}/);
  if (braceMatch) {
    try {
      return JSON.parse(braceMatch[0]) as T;
    } catch {
      // fall through
    }
  }
  const bracketMatch = text.match(/\[[\s\S]*\]/);
  if (bracketMatch) {
    try {
      return JSON.parse(bracketMatch[0]) as T;
    } catch {
      return null;
    }
  }
  return null;
}