catalyst-core 0.1.0-canary.9 → 0.2.0-beta.2

package/mcp_v2/lib/helpers.js

@@ -38,7 +38,8 @@ function versionOlderThan(a, b) {
 }
 
 /**
- * Walk up the directory tree looking for a package.json that depends on Catalyst.
+ * Walk up the directory tree looking for a package.json that depends on Catalyst,
+ * or the catalyst-core package source itself when developing this repo.
  * Returns { dir, pkg, catalystPackageName, catalystVersion, installedVersion, versionMeta } or null.
  */
 function findCatalystRoot() {
@@ -54,6 +55,26 @@ function findCatalystRoot() {
             if (fs.existsSync(pkgPath)) {
                 try {
                     const pkg = JSON.parse(fs.readFileSync(pkgPath, "utf8"))
+                    if (pkg.name === "catalyst-core") {
+                        const sourceVersion = pkg.version || null
+                        return {
+                            dir,
+                            pkg,
+                            catalystPackageName: "catalyst-core",
+                            catalystVersion: sourceVersion,
+                            version: sourceVersion,
+                            installedVersion: sourceVersion,
+                            notInstalled: false,
+                            isSourcePackage: true,
+                            versionMeta: {
+                                declaredRef: sourceVersion,
+                                isGithubRef: false,
+                                installedVersion: sourceVersion,
+                                parsed: parseVersion(sourceVersion),
+                            },
+                        }
+                    }
+
                     const deps = { ...pkg.dependencies, ...pkg.devDependencies }
                     const declaredRef = deps["catalyst-core"]
                     if (declaredRef) {
@@ -77,6 +98,7 @@ function findCatalystRoot() {
                             pkg,
                             catalystPackageName: "catalyst-core",
                             catalystVersion: declaredRef, // what package.json says (may be github ref)
+                            version: declaredRef,
                             installedVersion, // what's actually in node_modules e.g. "0.0.3-canary.3"
                             notInstalled: !installed,
                             versionMeta: {
@@ -101,21 +123,33 @@ function findCatalystRoot() {
  * File-system helpers scoped to a project root.
  */
 function makeProjectHelpers(root) {
+    function safeProjectPath(rel) {
+        if (typeof rel !== "string" || path.isAbsolute(rel)) return null
+        const normalized = path.normalize(rel)
+        if (normalized === ".." || normalized.startsWith(`..${path.sep}`)) return null
+        return `${root}${path.sep}${normalized}`
+    }
+
     function fileExists(rel) {
-        return fs.existsSync(path.join(root, rel))
+        const filePath = safeProjectPath(rel)
+        return Boolean(filePath && fs.existsSync(filePath))
     }
 
     function readJson(rel) {
+        const filePath = safeProjectPath(rel)
+        if (!filePath) return null
         try {
-            return JSON.parse(fs.readFileSync(path.join(root, rel), "utf8"))
+            return JSON.parse(fs.readFileSync(filePath, "utf8"))
         } catch {
             return null
         }
     }
 
     function readText(rel) {
+        const filePath = safeProjectPath(rel)
+        if (!filePath) return null
         try {
-            return fs.readFileSync(path.join(root, rel), "utf8")
+            return fs.readFileSync(filePath, "utf8")
         } catch {
             return null
         }
@@ -125,7 +159,6 @@ function makeProjectHelpers(root) {
      * Walk src/**\/*.{js,jsx,ts,tsx} and return relative paths that match pattern.
      */
     function grepSrc(pattern) {
-        const re = new RegExp(pattern)
         const matches = []
         function walk(dir) {
             let entries
@@ -136,12 +169,13 @@ function makeProjectHelpers(root) {
             }
             for (const e of entries) {
                 if (e.name === "node_modules" || e.name === ".git") continue
-                const full = path.join(dir, e.name)
+                const full = `${dir}${path.sep}${e.name}`
                 if (e.isDirectory()) {
                     walk(full)
                 } else if (/\.(js|jsx|ts|tsx)$/.test(e.name)) {
                     try {
-                        if (re.test(fs.readFileSync(full, "utf8"))) {
+                        const content = fs.readFileSync(full, "utf8")
+                        if (pattern.test(content)) {
                             matches.push(path.relative(root, full))
                         }
                     } catch {
@@ -150,7 +184,7 @@ function makeProjectHelpers(root) {
                 }
             }
         }
-        walk(path.join(root, "src"))
+        walk(`${root}${path.sep}src`)
         return matches
     }
 

package/mcp_v2/lib/seed.js

@@ -0,0 +1,88 @@
+"use strict"
+
+const fs = require("fs")
+
+/**
+ * Re-seed framework_knowledge (source='static') and known_errors from knowledge-base.json.
+ * Caller is responsible for rebuilding fk_fts afterwards (via knowledge.init).
+ *
+ * @param {import('better-sqlite3').Database} db
+ * @param {string} kbPath - absolute path to knowledge-base.json
+ * @param {?{ dir: string, pkg: { name?: string }, version: string }} projectInfo
+ *        When provided, also upserts the project_context row. Pass null at runtime sync
+ *        when project_context is already populated.
+ * @returns {{ knowledgeCount: number, errorCount: number }}
+ */
+function seedKnowledgeBase(db, kbPath, projectInfo = null) {
+    const kb = JSON.parse(fs.readFileSync(kbPath, "utf8"))
+    if (!Array.isArray(kb)) throw new Error("knowledge-base.json must be a JSON array")
+
+    const deleteKnowledge = db.prepare(`DELETE FROM framework_knowledge WHERE source = 'static'`)
+    const deleteErrors = db.prepare(`DELETE FROM known_errors`)
+
+    const insertKnowledge = db.prepare(`
+        INSERT INTO framework_knowledge (section, title, content, layer, source, tags, github_files)
+        VALUES (@section, @title, @content, @layer, @source, @tags, @github_files)
+    `)
+
+    const insertError = db.prepare(`
+        INSERT INTO known_errors (symptom, cause, fix, layer, tags)
+        VALUES (@symptom, @cause, @fix, @layer, @tags)
+    `)
+
+    let knowledgeCount = 0
+    let errorCount = 0
+
+    const seedAll = db.transaction(() => {
+        deleteKnowledge.run()
+        deleteErrors.run()
+        for (const entry of kb) {
+            if (entry.section === "known_errors") {
+                // known_errors entries encode symptom/cause/fix in content as:
+                // "Symptom: ... Cause: ... Fix: ..."
+                const content = entry.content || ""
+                const symptomMatch = content.match(/Symptom:\s*([^.]+\.?)/i)
+                const causeMatch = content.match(/Cause:\s*([^.]+\.?)/i)
+                const fixMatch = content.match(/Fix:\s*([\s\S]+)/i)
+                insertError.run({
+                    symptom: symptomMatch ? symptomMatch[1].trim() : entry.title,
+                    cause: causeMatch ? causeMatch[1].trim() : "",
+                    fix: fixMatch ? fixMatch[1].trim() : "",
+                    layer: entry.layer,
+                    tags: JSON.stringify(entry.tags || []),
+                })
+                errorCount++
+            } else {
+                insertKnowledge.run({
+                    section: entry.section,
+                    title: entry.title,
+                    content: entry.content,
+                    layer: entry.layer,
+                    source: "static",
+                    tags: JSON.stringify(entry.tags || []),
+                    github_files: entry.github_files ? JSON.stringify(entry.github_files) : null,
+                })
+                knowledgeCount++
+            }
+        }
+    })
+
+    seedAll()
+
+    if (projectInfo) {
+        db.prepare(
+            `
+            INSERT OR REPLACE INTO project_context (id, repo_path, package_name, catalyst_version, detected_at)
+            VALUES (1, @repo_path, @package_name, @catalyst_version, datetime('now'))
+        `
+        ).run({
+            repo_path: projectInfo.dir,
+            package_name: projectInfo.pkg.name || "unknown",
+            catalyst_version: projectInfo.version,
+        })
+    }
+
+    return { knowledgeCount, errorCount }
+}
+
+module.exports = { seedKnowledgeBase }

package/mcp_v2/mcp.js

@@ -25,6 +25,7 @@ const build = require("./tools/build")
 const tasks = require("./tools/tasks")
 const sync = require("./tools/sync")
 const knowledge = require("./tools/knowledge")
+const github = require("./tools/github")
 
 const MCP_DIR = __dirname
 const DB_PATH = path.join(MCP_DIR, "context.db")
@@ -87,6 +88,7 @@ build.init(db)
 tasks.init(db)
 sync.init(db)
 knowledge.init(db)
+github.init(projectInfo)
 
 // ── Intent classification (internal, not exposed as tool) ─────────────────────
 
@@ -97,6 +99,9 @@ const INTENT_PATTERNS = {
     guidance:
         /\b(what\s+is|what\s+are|how\s+does|how\s+do|explain|show\s+me|tell\s+me|hook|api|usage|example)\b/i,
     status: /status|done|complet|finish|check.*config|config.*check|what.*(left|remain|todo|next|pending)|how far|progress/i,
+    // feedback = wants to raise an issue or discussion on GitHub
+    feedback:
+        /\b(issue|bug\s+report|report\s+(a\s+)?bug|open\s+(an?\s+)?issue|create\s+(an?\s+)?issue|raise\s+(an?\s+)?issue|discussion|discuss|feature\s+request|proposal|suggest)\b/i,
     debug: /error|fail|broken|not work|crash|issue|bug|why|wrong/i,
     build: /build|compile|webpack|bundle|android|ios|platform/i,
     sync: /sync|update.*doc|fetch.*doc|latest.*doc/i,
@@ -110,6 +115,8 @@ const INTENT_NEXT_ACTION = {
     debug: "answer_only — provide debug guidance. Do NOT call create_task_plan.",
     build: "answer_only — explain build flow. Do NOT call create_task_plan.",
     sync: "answer_only — sync complete. Do NOT call create_task_plan.",
+    feedback:
+        "answer_only — run the GitHub issue workflow and show the created issue URL or markdown fallback. Do NOT call create_task_plan.",
     unknown: "answer_only — unclear intent. Return what you found. Do NOT call create_task_plan.",
 }
 
@@ -399,9 +406,9 @@ const TOOLS = [
         },
     },
     {
-        name: "sync_catalyst_docs",
+        name: "sync_knowledge_base",
         description:
-            "Use when the developer asks: 'sync docs', 'update framework knowledge', 'fetch latest catalyst docs'. Intent: sync. Fetches changelog and template diffs, updates the KB. Maintenance only — no task planning needed after.",
+            "Use when the developer asks: 'sync docs', 'update framework knowledge', 'fetch latest catalyst docs'. Intent: sync. Pulls the latest knowledge-base.json from tata1mg/catalyst-core@main and re-seeds the KB. Maintenance only — no task planning needed after.",
         inputSchema: {
             type: "object",
             properties: {
@@ -412,6 +419,144 @@ const TOOLS = [
             },
         },
     },
+    {
+        name: "create_github_issue",
+        description:
+            "Use when the developer wants to create, raise, report, preview, or publish a GitHub issue for catalyst-core, or after the LLM discovers that a problem is likely caused by catalyst-core framework behavior and the developer agrees to raise an issue. This single tool first asks the developer to select labels when labels are omitted. After labels are supplied, it gathers project context, renders the issue using the selected/suggested template, searches duplicates using duplicate_search_query when provided, supports preview with dry_run:true, publishes only when dry_run:false is explicitly passed, and falls back to a markdown draft on auth/network/API failure. Default to dry_run:true first; if label_selection_required is returned, ask the developer to select labels before collecting/rendering the rest of the issue. Intent: feedback.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                dry_run: {
+                    type: "boolean",
+                    description:
+                        "Defaults to true. When true, returns rendered preview, labels, duplicate candidates, and gathered context without publishing. Pass false only after explicit developer approval.",
+                },
+                project_path: {
+                    type: "string",
+                    description: "Path to the catalyst app root. Defaults to detected project root.",
+                },
+                title: {
+                    type: "string",
+                    description:
+                        "Short, descriptive issue title. E.g. 'RouterDataProvider fails on nested dynamic routes'.",
+                },
+                body: {
+                    type: "string",
+                    description:
+                        "Full issue description. Plain text is upgraded into the catalyst-core issue style; already structured markdown is preserved.",
+                },
+                summary: {
+                    type: "string",
+                    description: "Structured issue summary if body is not already composed.",
+                },
+                issue_template: {
+                    type: "string",
+                    enum: ["bug", "enhancement", "documentation", "dependencies", "question"],
+                    description:
+                        "Optional template to force. Use bug for broken behavior, enhancement for feature requests, documentation for docs/examples, dependencies for package updates, question for clarification.",
+                },
+                current_behavior: {
+                    type: "string",
+                    description: "What happens today.",
+                },
+                steps_to_reproduce: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Steps to reproduce the issue.",
+                },
+                expected_behavior: {
+                    type: "string",
+                    description: "What should happen.",
+                },
+                actual_behavior: {
+                    type: "string",
+                    description: "What actually happens.",
+                },
+                error_logs: {
+                    type: "string",
+                    description: "Error logs or stack traces.",
+                },
+                preflight_checklist: {
+                    type: "array",
+                    items: { type: "string" },
+                    description:
+                        "Optional checklist items for bug reports. Defaults to searched issues, single issue, and included repro/environment/logs.",
+                },
+                what_i_tried: {
+                    type: "string",
+                    description: "Troubleshooting already attempted.",
+                },
+                additional_information: {
+                    type: "string",
+                    description: "Additional context that does not fit the primary sections.",
+                },
+                related_issues: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Related GitHub issues or links.",
+                },
+                root_cause: {
+                    type: "string",
+                    description: "Technical notes or suspected root cause.",
+                },
+                proposed_fix: {
+                    type: "string",
+                    description: "Concrete suggested fix.",
+                },
+                optional_followups: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Follow-up cleanup, docs, or test suggestions.",
+                },
+                environment: {
+                    type: "object",
+                    description:
+                        "Environment details such as platform, OS, browser, device, Node/npm versions.",
+                },
+                labels: {
+                    type: "array",
+                    items: { type: "string" },
+                    description:
+                        "Required after the initial label-selection step. Valid catalyst-core labels: bug, dependencies, documentation, duplicate, enhancement, good first issue, help wanted, invalid, question, wontfix. If omitted, the tool returns label_selection_required and does not gather context, render a full preview, search duplicates, or publish.",
+                },
+                images: {
+                    type: "array",
+                    items: { type: "object" },
+                    description:
+                        "Optional image URLs or local image paths. URLs are embedded directly; local paths are preserved for fallback/manual upload.",
+                },
+                duplicate_search_query: {
+                    type: "string",
+                    description:
+                        "Optional focused search query for duplicate detection, supplied by the LLM from the issue context. Use the same query reviewed in the dry_run:true preview.",
+                },
+                duplicate_review_confirmed: {
+                    type: "boolean",
+                    description:
+                        "Set true only after a dry_run:true preview or blocked publish returned duplicate candidates, those candidates were shown to the developer, and the developer explicitly confirmed this issue is distinct and should still be published.",
+                },
+                sensitive_data_confirmed: {
+                    type: "boolean",
+                    description:
+                        "Set true only after a dry_run:true preview or blocked publish returned sensitive_data_review.required_before_publish=true, the config/sensitive-looking data warning and rendered issue preview were shown to the developer, and the developer explicitly confirmed this data may be posted to GitHub.",
+                },
+                duplicate_review_note: {
+                    type: "string",
+                    description:
+                        "Optional short note explaining why the issue is not a duplicate, after the developer confirms publishing.",
+                },
+                files: {
+                    type: "array",
+                    items: { type: "string" },
+                    description: "Optional explicit project files to include as context snippets.",
+                },
+                _query: {
+                    type: "string",
+                    description: "Original user query for intent classification.",
+                },
+            },
+        },
+    },
 ]
 
 // ── Tool handler dispatch ─────────────────────────────────────────────────────
@@ -427,8 +572,9 @@ const TOOL_HANDLERS = {
     update_task_step: tasks.handle_update_task_step,
     get_active_task: tasks.handle_get_active_task,
     close_task_plan: tasks.handle_close_task_plan,
-    sync_catalyst_docs: sync.handle_sync_catalyst_docs,
+    sync_knowledge_base: sync.handle_sync_knowledge_base,
     query_knowledge: knowledge.handle_query_knowledge,
+    create_github_issue: github.handle_create_github_issue,
 }
 
 // ── MCP JSON-RPC over stdio ───────────────────────────────────────────────────
@@ -496,7 +642,7 @@ rl.on("line", (line) => {
                                 text: JSON.stringify({
                                     error: "out_of_scope",
                                     message:
-                                        "This query is outside Catalyst MCP scope. MCP handles: conversion tracking, debugging, config validation, build flow, architecture, task planning, and doc sync.",
+                                        "This query is outside Catalyst MCP scope. MCP handles: conversion tracking, debugging, config validation, build flow, architecture, task planning, doc sync, and GitHub issue creation.",
                                 }),
                             },
                         ],

package/mcp_v2/setup.js

@@ -8,38 +8,22 @@
  * 2. Create context.db, run schema.sql
  * 3. Seed framework_knowledge from knowledge-base.json (static rows)
  * 4. Seed known_errors from knowledge-base.json known_errors section
- * 5. Run sync_catalyst_docs once (dynamic sitemap rows)
+ * 5. Build FTS index over static rows
+ * 6. Sync knowledge-base.json from tata1mg/catalyst-core@main (best-effort; warns but does not fail)
  */
 
 const fs = require("fs")
 const path = require("path")
 const Database = require("better-sqlite3")
-const https = require("https")
-const crypto = require("crypto")
+const sync = require("./tools/sync")
+const knowledge = require("./tools/knowledge")
+const { seedKnowledgeBase } = require("./lib/seed")
+const { findCatalystRoot } = require("./lib/helpers")
 
 const MCP_DIR = __dirname
 const DB_PATH = path.join(MCP_DIR, "context.db")
 const SCHEMA_PATH = path.join(MCP_DIR, "schema.sql")
 const KB_PATH = path.join(MCP_DIR, "knowledge-base.json")
-const SITEMAP_URL = "https://catalyst.1mg.com/public_docs/sitemap.xml"
-
-// ── 1. Find & validate catalyst project ──────────────────────────────────────
-
-function findCatalystRoot() {
-    let dir = process.cwd()
-    while (dir !== path.parse(dir).root) {
-        const pkgPath = path.join(dir, "package.json")
-        if (fs.existsSync(pkgPath)) {
-            const pkg = JSON.parse(fs.readFileSync(pkgPath, "utf8"))
-            const deps = { ...pkg.dependencies, ...pkg.devDependencies }
-            if (deps["catalyst-core"]) {
-                return { dir, pkg, catalystPackageName: "catalyst-core", version: deps["catalyst-core"] }
-            }
-        }
-        dir = path.dirname(dir)
-    }
-    return null
-}
 
 // ── 2. DB init ────────────────────────────────────────────────────────────────
 
@@ -52,180 +36,15 @@ function initDb() {
 
 // ── 3 & 4. Seed from knowledge-base.json ─────────────────────────────────────
 
-function seedKnowledgeBase(db, projectInfo) {
-    const kb = JSON.parse(fs.readFileSync(KB_PATH, "utf8"))
-
-    // DELETE existing static rows so re-runs reflect updated knowledge-base.json
-    db.prepare(`DELETE FROM framework_knowledge WHERE source = 'static'`).run()
-    db.prepare(`DELETE FROM known_errors`).run()
-
-    const insertKnowledge = db.prepare(`
-    INSERT INTO framework_knowledge (section, title, content, layer, source, tags, github_files)
-    VALUES (@section, @title, @content, @layer, @source, @tags, @github_files)
-  `)
-
-    const insertError = db.prepare(`
-    INSERT INTO known_errors (symptom, cause, fix, layer, tags)
-    VALUES (@symptom, @cause, @fix, @layer, @tags)
-  `)
-
-    let knowledgeCount = 0
-    let errorCount = 0
-
-    const seedAll = db.transaction(() => {
-        for (const entry of kb) {
-            if (entry.section === "known_errors") {
-                // known_errors entries encode symptom/cause/fix in content as:
-                // "Symptom: ... Cause: ... Fix: ..."
-                const content = entry.content || ""
-                const symptomMatch = content.match(/Symptom:\s*([^.]+\.?)/i)
-                const causeMatch = content.match(/Cause:\s*([^.]+\.?)/i)
-                const fixMatch = content.match(/Fix:\s*([\s\S]+)/i)
-                insertError.run({
-                    symptom: symptomMatch ? symptomMatch[1].trim() : entry.title,
-                    cause: causeMatch ? causeMatch[1].trim() : "",
-                    fix: fixMatch ? fixMatch[1].trim() : "",
-                    layer: entry.layer,
-                    tags: JSON.stringify(entry.tags || []),
-                })
-                errorCount++
-            } else {
-                insertKnowledge.run({
-                    section: entry.section,
-                    title: entry.title,
-                    content: entry.content,
-                    layer: entry.layer,
-                    source: "static",
-                    tags: JSON.stringify(entry.tags || []),
-                    github_files: entry.github_files ? JSON.stringify(entry.github_files) : null,
-                })
-                knowledgeCount++
-            }
-        }
-    })
-
-    seedAll()
+function seedFromLocalKb(db, projectInfo) {
+    const { knowledgeCount, errorCount } = seedKnowledgeBase(db, KB_PATH, projectInfo)
     console.log(`  ✓ Seeded ${knowledgeCount} knowledge entries`)
     console.log(`  ✓ Seeded ${errorCount} known_errors entries`)
-
-    // Store project context
-    db.prepare(
-        `
-    INSERT OR REPLACE INTO project_context (id, repo_path, package_name, catalyst_version, detected_at)
-    VALUES (1, @repo_path, @package_name, @catalyst_version, datetime('now'))
-  `
-    ).run({
-        repo_path: projectInfo.dir,
-        package_name: projectInfo.pkg.name || "unknown",
-        catalyst_version: projectInfo.version,
-    })
     console.log(
         `  ✓ Project context stored (${projectInfo.pkg.name}, ${projectInfo.catalystPackageName}@${projectInfo.version})`
     )
 }
 
-// ── 5. sync_catalyst_docs (initial run) ───────────────────────────────────────
-
-function fetchUrl(url) {
-    return new Promise((resolve, reject) => {
-        const mod = url.startsWith("https") ? https : require("http")
-        mod.get(url, (res) => {
-            if (res.statusCode >= 300 && res.statusCode < 400 && res.headers.location) {
-                return fetchUrl(res.headers.location).then(resolve).catch(reject)
-            }
-            let data = ""
-            res.on("data", (chunk) => (data += chunk))
-            res.on("end", () => resolve(data))
-        }).on("error", reject)
-    })
-}
-
-function parseSitemapUrls(xml) {
-    const matches = xml.match(/<loc>(.*?)<\/loc>/g) || []
-    return matches.map((m) => m.replace(/<\/?loc>/g, "").trim())
-}
-
-function stripHtml(html) {
-    // Very simple strip — good enough for catalyst docs
-    return html
-        .replace(/<script[\s\S]*?<\/script>/gi, "")
-        .replace(/<style[\s\S]*?<\/style>/gi, "")
-        .replace(/<[^>]+>/g, " ")
-        .replace(/\s+/g, " ")
-        .trim()
-}
-
-function contentHash(text) {
-    return crypto.createHash("sha256").update(text).digest("hex").slice(0, 16)
-}
-
-// Kept for manual/live sync flows; setup does not invoke it by default.
-// eslint-disable-next-line no-unused-vars
-async function syncCatalystDocs(db) {
-    console.log("\n  Fetching sitemap...")
-    let xml
-    try {
-        xml = await fetchUrl(SITEMAP_URL)
-    } catch (e) {
-        console.warn(`  ⚠ Could not fetch sitemap (${e.message}). Skipping live docs sync.`)
-        console.warn("    You can run sync later via the sync_catalyst_docs MCP tool.")
-        return
-    }
-
-    const urls = parseSitemapUrls(xml)
-    console.log(`  Found ${urls.length} URLs in sitemap`)
-
-    const insertKnowledge = db.prepare(`
-    INSERT INTO framework_knowledge (section, title, content, layer, source, tags, url)
-    VALUES (@section, @title, @content, @layer, 'sitemap', '[]', @url)
-  `)
-
-    const insertSnapshot = db.prepare(`
-    INSERT INTO doc_snapshots (url, content_hash, slot) VALUES (@url, @hash, 'curr')
-  `)
-
-    const insertLink = db.prepare(`
-    INSERT INTO linkage_map (url, knowledge_id) VALUES (@url, @knowledge_id)
-  `)
-
-    let synced = 0
-    let failed = 0
-
-    for (const url of urls) {
-        try {
-            const html = await fetchUrl(url)
-            const text = stripHtml(html)
-            const hash = contentHash(text)
-
-            // Extract a title from <title> tag if present
-            const titleMatch = html.match(/<title[^>]*>(.*?)<\/title>/i)
-            const title = titleMatch ? titleMatch[1].trim() : url.split("/").pop() || url
-
-            // Truncate content to 2000 chars to keep DB lean
-            const content = text.slice(0, 2000)
-
-            db.transaction(() => {
-                const row = insertKnowledge.run({
-                    section: "sitemap",
-                    title,
-                    content,
-                    layer: "Component",
-                    url,
-                })
-                insertSnapshot.run({ url, hash })
-                insertLink.run({ url, knowledge_id: row.lastInsertRowid })
-            })()
-
-            synced++
-            process.stdout.write(`\r  Synced ${synced}/${urls.length} pages...`)
-        } catch (e) {
-            failed++
-        }
-    }
-
-    console.log(`\n  ✓ Synced ${synced} pages (${failed} failed)`)
-}
-
 // ── Main ──────────────────────────────────────────────────────────────────────
 
 async function main() {
@@ -248,11 +67,29 @@ async function main() {
 
     // 3 & 4. Seed static knowledge
     console.log("\nSeeding knowledge-base.json...")
-    seedKnowledgeBase(db, projectInfo)
+    seedFromLocalKb(db, projectInfo)
 
-    // 5. Initial sitemap sync — disabled (slow, use sync_catalyst_docs MCP tool manually)
-    // console.log('\nRunning initial sync_catalyst_docs...');
-    // await syncCatalystDocs(db);
+    // 5. Build FTS index from the just-seeded static rows.
+    //    DROP first so re-runs of setup don't leave orphan rowids pointing at deleted rows.
+    db.exec(`DROP TABLE IF EXISTS fk_fts`)
+    knowledge.init(db)
+
+    // 6. Initial sync — pulls knowledge-base.json from tata1mg/catalyst-core@main.
+    //    Failures are warnings, not setup-blockers (network may be unavailable at install time).
+    console.log("\nSyncing knowledge-base.json from tata1mg/catalyst-core@main...")
+    sync.init(db)
+    try {
+        const result = await sync.handle_sync_knowledge_base({})
+        if (result.error) {
+            console.warn(`  ⚠ ${result.error}`)
+            console.warn("    Run sync_knowledge_base later via your MCP client to retry.")
+        } else {
+            console.log(`  ✓ ${result.message}`)
+        }
+    } catch (e) {
+        console.warn(`  ⚠ Sync threw: ${e.message}`)
+        console.warn("    Run sync_knowledge_base later via your MCP client to retry.")
+    }
 
     db.close()
     console.log("\n✓ Setup complete. MCP is ready.\n")