@skyphusion/search-mcp 0.4.0 → 0.5.0

package/README.md

@@ -6,7 +6,8 @@
 
 Open-source toolkit for [Cloudflare AI Search](https://developers.cloudflare.com/ai-search/):
 
-1. **MCP Worker** (`src/mcp.ts`) -- bearer-gated Streamable-HTTP MCP with a `search` tool for agents.
+1. **MCP Worker** (`src/mcp.ts`) -- bearer-gated Streamable-HTTP MCP for agents:
+   `search`, `list_repos`, `get_file`, `ask`, `corpus_status`, plus `corpus://` resources.
 2. **Query Worker** (`src/index.ts`) -- CORS + Turnstile + rate-limited `POST /ask` that streams answers for a browser widget.
 3. **Corpus sync** (`scripts/sync.mjs`, `scripts/sync-runner.mjs`) -- git-tracked sources to R2, with extension remapping so TypeScript, Dockerfiles, and other text AI Search would otherwise skip get indexed.
 
@@ -104,6 +105,20 @@ wrangler secret put TURNSTILE_SECRET   # optional; skips verification when unset
 
 `MCP_TOKEN` accepts a single token or comma-separated `name=token` pairs for per-consumer attribution in logs.
 
+### MCP tools (agents)
+
+| Tool | Role |
+| --- | --- |
+| `list_repos` | Exact repo names for filters (`CORPUS_REPOS` or R2 prefixes) |
+| `search` | Hybrid/keyword/vector chunks; optional `repos`, `path_prefix`, `min_score`, `rewrite` |
+| `get_file` | Capped R2 read by `repo` + `path` (needs `CORPUS` binding) |
+| `ask` | One grounded answer via chatCompletions + source list |
+| `corpus_status` | Last sync metadata (`_meta/corpus-status.json` from `sync.mjs`) |
+
+Resources: `corpus://catalog`, `corpus://skill`. Prefer **list_repos → search → get_file**; use **ask** when you want prose rather than raw chunks.
+
+Bind the same R2 bucket the sync uploads to as `CORPUS` on the MCP Worker so `get_file` and status work. Optional `CORPUS_REPOS` JSON pins the catalog without listing R2.
+
 ## Corpus sync
 
 ```sh

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@skyphusion/search-mcp",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "description": "Bearer-gated MCP server and streaming query Worker over Cloudflare AI Search, plus a git-to-R2 corpus sync toolkit.",
   "license": "AGPL-3.0-only",
   "author": "Conrad Rockenhaus <conrad@skyphusion.org>",
@@ -76,5 +76,8 @@
     "typescript": "^7.0.2",
     "vitest": "^4.1.10",
     "wrangler": "^4.110.0"
+  },
+  "overrides": {
+    "undici": "7.29.0"
   }
 }

package/scripts/sync.mjs

@@ -350,13 +350,36 @@ async function main() {
     ContinuationToken = res.IsTruncated ? res.NextContinuationToken : undefined;
   } while (ContinuationToken);
 
-  for (let i = 0; i < stale.length; i += 1000) {
-    const batch = stale.slice(i, i + 1000);
+  // Never prune operator/meta keys written for MCP corpus_status.
+  const stalePrune = stale.filter((o) => o.Key && !o.Key.startsWith("_meta/"));
+  for (let i = 0; i < stalePrune.length; i += 1000) {
+    const batch = stalePrune.slice(i, i + 1000);
     await s3.send(
       new DeleteObjectsCommand({ Bucket, Delete: { Objects: batch } }),
     );
   }
-  console.log(`Pruned ${stale.length} stale objects.`);
+  console.log(`Pruned ${stalePrune.length} stale objects.`);
+
+  // MCP corpus_status + list_repos freshness (search-mcp agent tooling).
+  const status = {
+    ts: Date.now(),
+    ok: true,
+    target: targetName,
+    bucket: target.bucket,
+    instance: target.instance,
+    objectCount: plan.length,
+    repos: [...target.repos],
+    note: "Written by scripts/sync.mjs after successful upload+prune.",
+  };
+  await s3.send(
+    new PutObjectCommand({
+      Bucket,
+      Key: "_meta/corpus-status.json",
+      Body: Buffer.from(JSON.stringify(status, null, 2), "utf8"),
+      ContentType: "application/json",
+    }),
+  );
+  console.log("Wrote _meta/corpus-status.json for MCP corpus_status.");
   console.log("Done. AI Search will re-index on its next sync.");
 }